lispguide.xml

<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="styleguide.xsl"?>
<GUIDE title="Google Common Lisp Style Guide">


<p align="right">

Revision 1.17
</p>


<address>
Robert Brown
</address>
<address>
  <a HREF="mailto:tunes@google.com">Fran&#231;ois-Ren&#233; Rideau</a>
</address>

<address>
   In memoriam Dan Weinreb
</address>

<p align="center">
<cite>Patterns mean "I have run out of language."</cite> &#8212; Rich Hickey
</p>


<OVERVIEW>
<CATEGORY title="Important Note">
  <STYLEPOINT title="Note: Displaying Hidden Details in this Guide">
     <SUMMARY>
       This style guide contains many details
       that are initially hidden from view.
       They are marked by the triangle icon, which you see here on your left.
       Click it now. You should see "Hooray" appear below.
     </SUMMARY>
     <BODY>
       <p>
        Hooray!  Now you know you can expand points to get more details.
        Alternatively, there's an "expand all" at the top of this document.
       </p>
     </BODY>
  </STYLEPOINT>
</CATEGORY>
<CATEGORY title="Background">
  <p>
    Common Lisp is a powerful multiparadigm programming language.
    With great power comes great responsibility.
  </p>
  <p>
    This guide recommends formatting and stylistic choices
    designed to make your code easier for other people to understand.
    For those internal applications and free software libraries that
    we develop at Google,
    you should keep within these guidelines when making changes.
    Note however that each project has its own rules and customs
    that complement or override these general guidelines;
    the speed-oriented QPX low fare search engine notably
    has a very different style and feel from the QRes reservation system.
  </p>
  <p>
    If you're writing Common Lisp code outside Google,
    we invite you to consider these guidelines.
    You may find some of them useful
    where they don't conflict with other priorities you have.
    We welcome remarks and constructive feedback
    on how to improve our guide, and
    on what alternate styles work for you and why.
  </p>
  
  <p>
    This guide is not a Common Lisp tutorial.
    For basic information about the language, please consult
    <a HREF="http://www.gigamonkeys.com/book/">Practical Common Lisp</a>.
    For a language reference, please consult the
    <a HREF="http://www.lispworks.com/documentation/HyperSpec/Front/index.htm">Common Lisp HyperSpec</a>.
    For more detailed style guidance, take a look at Peter Norvig and Kent Pitman's
    <a HREF="http://norvig.com/luv-slides.ps">style guide</a>.
  </p>
</CATEGORY>
</OVERVIEW>
<CATEGORY title="Meta-Guide">
  <STYLEPOINT title="Must, Should, May, or Not">
    <SUMMARY>
      Each guideline's level of importance is indicated
      by use of the following keywords and phrases, adapted from
      <a href="http://www.ietf.org/rfc/rfc2119.txt">RFC 2119</a>.
    </SUMMARY>
    <BODY>
      <table>
        <tr>
          <th valign="top">MUST</th>
          <td>
            <p>
              This word, or the terms "REQUIRED" or "SHALL",
              means that the guideline is an absolute requirement.
              You must ask permission to violate a MUST.
            </p>
          </td>
        </tr>
        <tr>
          <th valign="top">MUST NOT</th>
          <td>
            <p>
              This phrase, or the phrase "SHALL NOT",
              means that the guideline is an absolute prohibition.
              You must ask permission to violate a MUST NOT.
            </p>
          </td>
        </tr>
        <tr>
          <th valign="top">SHOULD</th>
          <td>
            <p>
              This word, or the adjective "RECOMMENDED", means that
              there may exist valid reasons in particular circumstances
              to ignore the demands of the guideline, but
              the full implications must be understood and carefully weighed
              before choosing a different course.
              You must ask forgiveness for violating a SHOULD.
            </p>
          </td>
        </tr>
        <tr>
          <th valign="top">SHOULD NOT</th>
          <td>
            <p>
              This phrase, or the phrase "NOT RECOMMENDED", means that
              there may exist valid reasons in particular circumstances
              to ignore the prohibitions of this guideline, but
              the full implications should be understood and carefully weighed
              before choosing a different course.
              You must ask forgiveness for violating a SHOULD NOT.
            </p>
          </td>
        </tr>
        <tr>
          <th valign="top">MAY</th>
          <td>
            <p>
              This word, or the adjective "OPTIONAL",
              means that an item is truly optional.
            </p>
          </td>
        </tr>
      </table>
      <p>
        Unlike RFCs, we don't capitalize every instance of one of the above
        keywords when it is used.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Permission and Forgiveness">
    <SUMMARY>
      There are cases where transgression of some of these rules
      is useful or even necessary.
      In some cases, you must seek permission or obtain forgiveness
      from the proper people.
    </SUMMARY>
    <BODY>
      <p>
        Permission comes from the owners of your project.
      </p>
      
      <p>
        Forgiveness is requested in a comment
        near the point of guideline violation,
        and is granted by your code reviewer.
        The original comment should be signed by you, and
        the reviewer should add a signed approval to the comment at review time.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Conventions">
    <SUMMARY>
      You MUST follow conventions. They are not optional.
    </SUMMARY>
    <BODY>
      <p>
        Some of these guidelines are motivated by universal principles of good programming.
        Some guidelines are motivated by technical peculiarities of Common Lisp.
        Some guidelines where once motivated by a technical reason,
        but the guideline remained after the reason subsided.
        Some guidelines, such those about as comments and indentation,
        are based purely on convention, rather than on clear technical merit.
        Whatever the case may be, you must still follow these guidelines,
        as well as other conventional guidelines
        that have not been formalized in this document.
      </p>
      <p>
        You MUST follow conventions.
        They are important for readability.
        When conventions are followed by default,
        violations of the convention are a signal
        that something notable is happening and deserves attention.
        When conventions are systematically violated,
        violations of the convention are a distracting noise
        that needs to be ignored.
      </p>
      <p>
        Conventional guidelines <em>are</em> indoctrination.
        Their purpose is to make you follow the mores of the community,
        
        so you can more effectively cooperate with existing members.
        It is still useful to distinguish the parts that are technically motivated
        from the parts that are mere conventions,
        so you know when best to defy conventions for good effect,
        and when not to fall into the pitfalls that the conventions are there to help avoid.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Old Code">
    <p>
      A lot of our code was written before these guidelines existed.
      You should fix violations as you encounter them
      in the course of your normal coding.
      You must not fix violations en masse
      without warning other developers and coordinating with them,
      so as not to make the merging of large branches
      more difficult than it already is.
    </p>
  </STYLEPOINT>
  <STYLEPOINT title="Future Topics">
    <SUMMARY>
      There are many topics for additional standardization
      not covered by current version of this document,
      but deferred to future versions.
    </SUMMARY>
    <BODY>
      <ul>
        <li>
          File and directory structure
        </li>
        <li>
          Packages and modularity
        </li>
        <li>
          Threads and locking
        </li>
        <li>
          How to add configurable components
        </li>
        <li>
          CLOS style: initforms, slot and accessor names, etc.
        </li>
        <li>
          Recommendations on max number of slots per class.
        </li>
        <li>
          More concrete examples of good code:
          <ul>
            <li>
              exceptions
            </li>
            <li>
              transactions, with retry
            </li>
            <li>
              XML
            </li>
            <li>
              typing
            </li>
            <li>
              encapsulation / abstraction
            </li>
            <li>
              class and slot names
            </li>
            <li>
              etc.
            </li>
          </ul>
        </li>
        <li>
          When (not) to use conditional compilation:
          <ul>
            <li>
              modifying the product
            </li>
            <li>
              conditional debugging/console output/etc.
            </li>
            <li>
              "temporarily" commenting-out blocks of code
            </li>
            <li>
              etc.
            </li>
          </ul>
        </li>
      </ul>
    </BODY>
  </STYLEPOINT>
  </CATEGORY>
<CATEGORY title="General Guidelines">
  <STYLEPOINT title="Principles">
    <SUMMARY>
      There are some basic principles for team software development
      that every developer must keep in mind.
      Whenever the detailed guidelines are inadequate, confusing or contradictory,
      refer back to these principles for guidance:
      <ul>
        <li>
          Every developer's code must be easy for another developer
          to read, understand, and modify
          &#8212; even if the first developer isn't around to explain it.
          (This is the "hit by a truck" principle.)
        </li>
        <li>
          Everybody's code should look the same.
          Ideally, there should be no way to look at lines of code
          and recognize it as "Fred's code" by its style.
        </li>
        <li>
          Don't be "clever" &#8212;
          do the simplest thing that could possibly work properly.
        </li>
        <li>
          Be precise.
        </li>
        <li>
          Be concise.
        </li>
        <li>
          KISS &#8212; Keep It Simple, Stupid.
        </li>
        <li>
          Use the smallest hammer for the job.
        </li>
        <li>
          Use common sense.
        </li>
        <li>
          Keep related code together.
          Minimize the amount of jumping around
          someone has to do to understand an area of code.
        </li>
      </ul>
    </SUMMARY>
    <BODY>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Priorities">
    <SUMMARY>
      <p>
        When making decisions about how to write a given piece of
        code, aim for the following -ilities in this priority order:
      </p>
      <ul>
        <li>
          Usability by the customer
        </li>
        <li>
          Debuggability/Testability
        </li>
        <li>
          Readability/Comprehensibility
        </li>
        <li>
          Extensibility/Modifiability
        </li>
        <li>
          Efficiency (of the Lisp code at runtime)
        </li>
      </ul>
    </SUMMARY>
    <BODY>
      <p>
        Most of these are obvious.
      </p>
      <p>
        Usability by the customer means that the system has to do what the
        customer requires; it has to handle the customer's transaction
        volumes, uptime requirements; etc.
      </p>
      <p>
        For the Lisp efficiency point,
        given two options of equivalent complexity,
        pick the one that performs better.
        (This is often the same as the one that conses less,
        i.e. allocates less storage from the heap.)
      </p>
      <p>
        Given two options where one is more complex than the other,
        pick the simpler option and revisit the decision only if
        profiling shows it to be a performance bottleneck.
      </p>
      <p>
        However, avoid premature optimization.
        Don't add complexity to speed up something that runs rarely,
        since in the long run, it matters less whether such code is fast.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Architecture">
    <SUMMARY>
      To build code that is robust and maintainable,
      it matters a lot how the code is divided into components,
      how these components communicate,
      how changes propagate as they evolve,
      and more importantly
      how the programmers who develop these components communicate
      as these components evolve.
    </SUMMARY>
    <BODY>
      <p>
        If your work affects other groups, might be reusable across groups,
        adds new components, has an impact on other groups
        (including QA or Ops), or otherwise isn't purely local,
        you must write it up using at least a couple of paragraphs,
        and get a design approval from the other parties involved
        before starting to write code &#8212; or be ready to scratch what you have
        when they object.
      </p>
      <p>
        If you don't know or don't care about these issues,
        ask someone who does.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Using Libraries">
    <SUMMARY>
      Often, the smallest hammer is to use an existing library.
      Or one that doesn't exist yet.
      In such cases, you are encouraged to use or develop such a library,
      but you must take appropriate precautions.
    </SUMMARY>
    <BODY>
      <ul>
        <li>
          You MUST NOT start a new library
          unless you established that none is already available
          that can be fixed or completed into becoming what you need.
          That's a rule against the NIH syndrome ("Not Invented Here"),
          which is particularly strong amongst Lisp hackers.
        </li>
        <li>
          Whichever library, old or new, you pick, you MUST get permission
          to incorporate third-party code into the code base.
          You must discuss the use of such library
          in the appropriate mailing-list,
          and have your code reviewed by people knowledgeable in the domain
          and/or the Lisp library ecosystem (if any).
          Please be ready to argue why this particular solution makes sense
          as compared to other available libraries.
        </li>
        <li>
          Some libraries are distributed under licenses not compatible
          with the software you're writing, and
          must not be considered available for use.
          Be aware of these issues, or consult with people who are.
        </li>
      </ul>
      
      <p>
        Note that if you have a "clever" implementation trick,
        and your trick really is clever,
        then you must definitely not include it in business specific code;
        but it may have its place in an open-source library used by the code.
        If your idea is not general purpose enough to have any users
        beyond your regular business users,
        then it is definitely either not clever enough or way too clever,
        and in either case does not belong in the code.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Open-Sourcing Code">
    <SUMMARY>
      <p>
        If you write a general-purpose library,
        or modify an existing open-source library,
        you are encouraged to publish the result
        separate from your main project and then
        have your project import it like any other open-source library.
      </p>
      
    </SUMMARY>
    <BODY>
      <p>
        Use your judgment to distinguish
        general-purpose versus business-specific code,
        and open-source the general-purpose parts,
        while keeping the business-specific parts a trade secret.
      </p>
      
      <p>
        Open-Sourcing code has many advantages,
        including being able to leverage third parties for development,
        letting the development of features be user-directed,
        and keeping you honest with respect to code quality.
        Whatever code you write, you will have to maintain anyway,
        and make sure its quality is high enough to sustain use in production.
        There should therefore be no additional burden to Open-Sourcing,
        even of code that (at least initially)
        is not directly usable by third parties.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Development Process">
    <SUMMARY>
      Development process is outside the scope of this document.
      However, developers should remember at least these bits:
      get reviewed, write tests, eliminate warnings, run tests, avoid mass-changes.
    </SUMMARY>
    <BODY>
      <p>
      </p>
      <ul>
        <li>
          All code changes must be reviewed.
          You should expect that your code will be reviewed by other hackers,
          and that you will be assigned other hackers' code to review.
          Part of the review criteria will be that code obeys
          the coding standards in this document.
        </li>
        <li>
          You must write and check-in tests for new code that you write and old bugs you fix.
          There must be a unit test for every API function,
          and any previously failing case.
          Your work is not truly done until this activity is complete.
          Estimating tasks must include the time it takes to produce such tests.
        </li>
        <li>
          Your code must compile
          without any compilation error or warning messages whatsoever.
          If the compiler issues warnings that should be ignored,
          muffle those warnings using the
          <code>xcvb-driver:with-controlled-compiler-conditions</code> and
          <code>xcvb-driver:*uninteresting-conditions*</code>
          framework (also available as <code>asdf-condition-control</code>),
          either around the entire project, or around individual files
          (using asdf's <code>:around-compile</code> hooks).
        </li>
        <li>
          All code should be checked in an appropriate source control system,
          in a way that allows for complete reproducibility of
          build, test and execution of
          the code that is, has been or may be deployed.
        </li>
        <li>
          You must run the "precheckin" tests, and each component must pass
          its unit tests successfully before you commit any code.
        </li>
        <li>
          You should incorporate code coverage into your testing process.
          Tests are not sufficient
          if they do not cover all new and updated code;
          code that for whatever reason cannot be included in coverage results
          should be clearly marked as such including the reason.
        </li>
        <li>
          Many people develop on branches.
          You must get permission to undertake mass-changes
          (e.g. mass reindentations)
          so that we can coordinate in advance,
          and give branch residents time to get back on the mainline
        </li>
      </ul>
    </BODY>
  </STYLEPOINT>
</CATEGORY>
<CATEGORY title="Formatting">
  <STYLEPOINT title="Spelling">
    <SUMMARY>
      <p>
        You must use correct spelling in your comments,
        and most importantly in your identifiers.
      </p>
      <p>
        When several correct spellings exist (including American vs English),
        and there isn't a consensus amongst developers as which to use,
        you should choose the shorter spelling.
      </p>
      <p>
        You must avoid using abbreviations for words,
        unless it's a word that is used very frequently,
        in which case you must use
        the <em>same</em> abbreviation <em>consistently</em>.
      </p>
    </SUMMARY>
    <BODY>
      <p>
        If you're not sure, consult a dictionary,
        Google for alternative spellings,
        or ask a local expert.
      </p>
      <p>
        Here are examples of choosing the correct spelling:
      </p>
      <ul>
        <li>
          Use "complimentary" in the sense of a meal or beverage
          that is not paid for by the recipient, not "complementary".
        </li>
        <li>
          Use "existent" and "nonexistent", not "existant".
          Use "existence", not "existance".
        </li>
        <li>
          Use "hierarchy" not "heirarchy".
        </li>
        <li>
          Use "precede" not "preceed".
        </li>
        <li>
          Use "weird", not "wierd".
        </li>
      </ul>
      <p>
        Here are examples of choosing the shorter spelling:
      </p>
      <ul>
        <li>
          Use "canceled", not "cancelled"
        </li>
        <li>
          Use "queuing", not "queueing".
        </li>
        <li>
          Use "signaled", not "signalled";
        </li>
        <li>
          Use "traveled", not "travelled".
        </li>
        <li>
          Use "aluminum", not "aluminium"
        </li>
        <li>
          Use "oriented", not "orientated"
        </li>
        <li>
          Use "color", not "colour"
        </li>
        <li>
          Use "behavior", not "behaviour"
        </li>
      </ul>
      <p>
        Make appropriate exceptions for industry standard nomenclature/jargon,
        including plain misspellings.
        For instance:
      </p>
      <ul>
        <li>
          Use "referer", not "referrer", in the context of the HTTP protocol.
        </li>
      </ul>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Line length">
    <SUMMARY>
      You should format source code so that no line is longer than 100 characters.
    </SUMMARY>
    <BODY>
      <p>
        Some line length restriction is better than none at all.
        While old text terminals used to make 80 columns the standard,
        these days, allowing 100 columns seems better,
        since good style encourages the use of
        descriptive variables and function names.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Indentation">
    <SUMMARY>
      <p>
        Indent your code the way a properly configured GNU Emacs does.
      </p>
      <p>
        Indent carefully to make the code easier to understand.
      </p>
    </SUMMARY>
    <BODY>
      <p>
        By default, GNU Emacs does an excellent job indenting Common Lisp code.
        It can be taught how to indent new defining forms
        and special rules for domain specific languages.
        Each project may have some file to customize indentation; use it.
      </p>
      
      
      <p>
        Use indentation to make complex function applications easier to read.
        When an application does not fit on one line
        or the function takes many arguments,
        consider inserting newlines between the arguments
        so that each one is on a separate line.
        However, do not insert newlines in a way that makes it hard to tell
        how many arguments the function takes
        or where an argument form starts and ends.
      </p>
      <BAD_CODE_SNIPPET>
        ;; Bad
        (do-something first-argument second-argument (lambda (x)
            (frob x)) fourth-argument last-argument)
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        ;; Better
        (do-something first-argument
                      second-argument
                      #'(lambda (x) (frob x))
                      fourth-argument
                      last-argument)
      </CODE_SNIPPET>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="File Header">
    <SUMMARY>
      <p>
        You must include maintainership and other important information
        at the top of each source file.
      </p>
      <p>
        You should not include a copyright statement in source files.
      </p>
    </SUMMARY>
    <BODY>
      <p>
        Every source file may begin with a one-line description of the file.
      </p>
      <p>
        After that optional description,
        every source file may prominently include a statement
        about who originally wrote the code,
        made major changes, and/or is the current owner/maintainer.
        This makes it easier for hackers to locate
        whom to ask questions about the code,
        or to identify that no one is left to reply to such inquiries.
        However, consider that the information
        is not very helpful if it is not maintained;
        unless it brings information
        that cannot be easily extracted from source control,
        it is better skipped.
      </p>
      <p>
        After that optional statement, every file should follow with
        a brief explanation of what the file contains.
      </p>
      <p>
        After that explanation, every file should start the code itself with an
        <code>(in-package :<em>package-name</em>)</code> form.
      </p>
      <p>
        After that <code>in-package</code> form,
        every file should follow with any file-specific
        <code>(declaim (optimize ...))</code> declaration
        that is not covered by an asdf <code>:around-compile</code> hook.
      </p>
      <CODE_SNIPPET>
        ;;;; Author: brown (Robert Brown)

        ;;;; Variable length encoding for integers and floating point numbers.

        (in-package #:varint)
        (declaim #.*optimize-default*)
      </CODE_SNIPPET>
      <p>
        You should not include copyright information in individual source code files.
        An exception is made for files meant to be disseminated as standalone.
      </p>
      <p>
        Each project or library has a single file specifying its license.
        Absence of a <tt>LICENSE</tt> or <tt>COPYING</tt> file
        means the project is proprietary code.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Vertical white space">
    <SUMMARY>
      Vertical white space: one blank line between top-level forms.
    </SUMMARY>
    <BODY>
      <p>
        You should include one blank line between top-level forms,
        such as function definitions.
        Exceptionally, blank lines can be omitted
        between simple, closely related defining forms of the same kind,
        such as a group of related type declarations or constant definitions.
      </p>
      <CODE_SNIPPET>
        (defconstant +mix32+ #x12b9b0a1 "pi, an arbitrary number")
        (defconstant +mix64+ #x2b992ddfa23249d6 "more digits of pi")

        (defconstant +golden-ratio32+ #x9e3779b9 "the golden ratio")
        (defconstant +golden-ratio64+ #xe08c1d668b756f82 "more digits of the golden ratio")

        (defmacro incf32 (x y)
          "Like INCF, but for integers modulo 2**32"
          `(setf ,x (logand (+ ,x ,y) #xffffffff)))
        (defmacro incf64 (x y)
          "Like INCF, but for integers modulo 2**64"
          `(setf ,x (logand (+ ,x ,y) #xffffffffffffffff)))
      </CODE_SNIPPET>
      <p>
        Blank lines can be used to separate parts of a complicated function.
        Generally, however, you should break a large function into smaller ones
        instead of trying to make it more readable by adding vertical space.
        If you can't, you should document with a <code>;;</code> comment
        what each of the separated parts of the function does.
      </p>
      <p>
        Every top-level form
        should be fewer than 61 lines long,
        including comments but excluding the documentation string.
        This applies to each of the forms in an <code>eval-when</code>,
        rather than to the <code>eval-when</code> itself.
        Additionally, <code>defpackage</code> forms may be longer,
        since they may include long lists of symbols.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Horizontal white space">
    <SUMMARY>
      Horizontal white space: none around parentheses. No tabs.
    </SUMMARY>
    <BODY>
      <p>
        You must not include extra horizontal whitespace
        before or after parentheses or around symbols.
      </p>
      <p>
        You must not place right parentheses by themselves on a line.
        A set of consecutive trailing parentheses must appear on the same line.
      </p>
      <BAD_CODE_SNIPPET>
        ;; Very Bad
        ( defun factorial ( limit )
          ( let (( product 1 ))
            ( loop for i from 1 upto limit
                  do (setf product ( * product i ) ) )
            product
          )
        )
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        ;; Better
        (defun factorial (limit)
          (let ((product 1))
            (loop for i from 1 upto limit
                  do (setf product (* product i)))
            product))
      </CODE_SNIPPET>
      <p>
        You should use only one space between forms.
      </p>
      <p>
        You should not use spaces to vertically align forms
        in the middle of consecutive lines.
        An exception is made when the code possesses
        an important yet otherwise not visible symmetry
        that you want to emphasize.
      </p>
      <BAD_CODE_SNIPPET>
        ;; Bad
        (let* ((low    1)
               (high   2)
               (sum    (+ (* low low) (* high high))))
          ...)
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        ;; Better
        (let* ((low 1)
               (high 2)
               (sum (+ (* low low) (* high high))))
          ...))
      </CODE_SNIPPET>
      <p>
        You must align nested forms if they occur across more than one line.
      </p>
      <BAD_CODE_SNIPPET>
        ;; Bad
        (defun munge (a b c)
        (* (+ a b)
        c))
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        ;; Better
        (defun munge (a b c)
          (* (+ a b)
             c))
      </CODE_SNIPPET>
      <p>
        The convention is that the body of a binding form
        is indented two spaces after the form.
        Any binding data before the body is usually indented four spaces.
        Arguments to a function call are aligned with the first argument;
        if the first argument is on its own line,
        it is aligned with the function name.
      </p>
      <CODE_SNIPPET>
        (multiple-value-bind (a b c d)
            (function-returning-four-values x y)
          (declare (ignore c))
          (something-using a)
          (also-using b d))
      </CODE_SNIPPET>
      <p>
        An exception to the rule against lonely parentheses
        is made for an <code>eval-when</code> form around several definitions;
        in this case, include a comment <code>; eval-when</code>
        after the closing parenthesis.
      </p>
      <p>
        You must set your editor to
        avoid inserting tab characters in the files you edit.
        Tabs cause confusion when editors disagree
        on how many spaces they represent.
        In Emacs, do <code>(setq-default indent-tabs-mode nil)</code>.
      </p>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Documentation">
  <STYLEPOINT title="Document everything">
    <SUMMARY>
      You should use document strings on all visible functions
      to explain how to use your code.
    </SUMMARY>
    <BODY>
      <p>
        Unless some bit of code is painfully self-explanatory,
        document it with a documentation string (also known as docstring).
      </p>
      <p>
        Documentation strings are destined to be read
        by the programmers who use your code.
        They can be extracted from functions, types, classes, variables and macros,
        and displayed by programming tools, such as IDEs,
        or by REPL queries such as <code>(describe 'foo)</code>;
        web-based documentation or other reference works
        can be created based on them.
        Documentation strings are thus the perfect locus to document your API.
        They should describe how to use the code
        (including what pitfalls to avoid),
        as opposed to how the code works (and where more work is needed),
        which is what you'll put in comments.
      </p>
      <p>
        Supply a documentation string when defining
        top-level functions, types, classes, variables and macros.
        Generally, add a documentation string wherever the language allows.
      </p>
      <p>
        For functions, the docstring should describe the function's contract:
        what the function does,
        what the arguments mean,
        what values are returned,
        what conditions the function can signal.
        It should be expressed at the appropriate level of abstraction,
        explaining the intended meaning rather than, say, just the syntax.
        In documentation strings, capitalize the names of Lisp symbols,
        such as function arguments.
        For example, "The value of LENGTH should be an integer."
      </p>
      <CODE_SNIPPET>
        (defun small-prime-number-p (n)
          "Return T if N, an integer, is a prime number. Otherwise, return NIL."
          (cond ((or (&lt; n 2))
                 nil)
                ((= n 2)
                 t)
                ((divisorp 2 n)
                 nil)
                (t
                 (loop for i from 3 upto (sqrt n) by 2
                       never (divisorp i n)))))
      </CODE_SNIPPET>
      <CODE_SNIPPET>
        (defgeneric table-clear (table)
          (:documentation
            "Like clrhash, empties the TABLE of all
            associations, and returns the table itself."))
      </CODE_SNIPPET>
      <p>
        A long docstring may usefully
        begin with a short, single-sentence summary,
        followed by the larger body of the docstring.
      </p>
      <p>
        When the name of a type is used,
        the symbol may be quoted by surrounding it with
        a back quote at the beginning and a single quote at the end.
        Emacs will highlight the type, and the highlighting serves
        as a cue to the reader that <kbd>M-.</kbd>
        will lead to the symbol's definition.
      </p>
      <CODE_SNIPPET>
        (defun bag-tag-expected-itinerary (bag-tag)
          "Return a list of `legacy-pnr-pax-segment' objects representing
          the expected itinerary of the `bag-tag' object, BAG-TAG."
          ...)
      </CODE_SNIPPET>
      <p>
        Every method of a generic function should be independently documented
        when the specialization affects what the method does,
        beyond what is described in its generic function's docstring.
      </p>
      <p>
        When you fix a bug,
        consider whether what the fixed code does is obviously correct or not;
        if not, you must add a comment explaining
        the reason for the code in terms of fixing the bug.
        Adding the bug number, if any, is also recommended.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Comment semicolons">
    <SUMMARY>
      You must use the appropriate number of semicolons to introduce comments.
    </SUMMARY>
    <BODY>
      <p>
        Comments are explanations to the future maintainers of the code.
        Even if you're the only person who will ever see and touch the code,
        even if you're either immortal and never going to quit,
        or unconcerned with what happens after you leave
        (and have your code self-destruct in such an eventuality),
        you may find it useful to comment your code.
        Indeed, by the time you revisit your code,
        weeks, months or years later,
        you will find yourself a different person from the one who wrote it,
        and you will be grateful to that previous self
        for making the code readable.
      </p>
      <p>
        You must comment anything complicated
        so that the next developer can understand what's going on.
        (Again, the "hit by a truck" principle.)
      </p>
      <p>
        Also use comments as a way to guide those who read the code,
        so they know what to find where.
      </p>
      <ul>
        <li>
          File headers and important comments
          that apply to large sections of code in a source file
          should begin with four semicolons.
        </li>
        <li>
          You should use three semicolons
          to begin comments that apply to just
          one top-level form or small group of top-level forms.
        </li>
        <li>
          Inside a top-level form, you should use two semicolons
          to begin a comment if it appears between lines.
        </li>
        <li>
          You should use one semicolon if it is a parenthetical remark
          and occurs at the end of a line.
          You should use spaces to separate the comment
          from the code it refers to so the comment stands out.
          You should try to vertically align
          consecutive related end-of-line comments.
        </li>
      </ul>
      <CODE_SNIPPET>
        ;;;; project-euler.lisp
        ;;;; File-level comments or comments for large sections of code.

        ;;; Problems are described in more detail here:  http://projecteuler.net/

        ;;; Divisibility
        ;;; Comments that describe a group of definitions.

        (defun divisorp (d n)
          (zerop (mod n d)))

        (defun proper-divisors (n)
          ...)

        (defun divisors (n)
          (cons n (proper-divisors n)))

        ;;; Prime numbers

        (defun small-prime-number-p (n)
          (cond ((or (&lt; n 2))
                 nil)
                ((= n 2)   ; parenthetical remark here
                 t)        ; continuation of the remark
                ((divisorp 2 n)
                 nil)  ; different remark
                ;; Comment that applies to a section of code.
                (t
                 (loop for i from 3 upto (sqrt n) by 2
                       never (divisorp i n)))))
      </CODE_SNIPPET>
      <p>
        You should include a space between the semicolon and the text of the comment.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Grammar and punctuation">
    <SUMMARY>
      You should punctuate documentation correctly.
    </SUMMARY>
    <BODY>
      <p>
        When a comment is a full sentence,
        you should capitalize the initial letter of the first word
        and end the comment with a period.
        In general, you should use correct punctuation.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Attention Required">
    <SUMMARY>
      You must follow the convention of using <code>---</code>
      for comments requiring special attention,
      including unobvious tricks, TODO items, questions, breakage, danger.
    </SUMMARY>
    <BODY>
      <ul>
        <li>
          <code>;--- </code> prefixes a cautionary comment,
          e.g. explaining why the code in question is particularly
          tricky, delicate, or non-obvious.
        </li>
        <li>
          <code>;---??? </code> prefixes a serious question
          which needs to be resolved soon,
          by fixing either the code or its documentation.
        </li>
        <li>
          <code>;---!!! </code> identifies code which is broken,
          but which for some reason you cannot fix at this time.
          You should not use this often for new code.
        </li>
        <li>
          <code>;---*** </code> identifies active <em>DANGER</em>,
          for instance where important functionality is stubbed out,
          or a large design issue remains unresolved.
          Anything so marked must be fixed
          before code is rolled into production.
        </li>
      </ul>
      <p>
        You must sign and date
        any of the above "requiring further attention" comments
        (but not mere cautionary explanations).
        You must use the
        <a HREF="http://www.w3.org/TR/NOTE-datetime">YYYY-MM-DD</a>
        format for dates to make automated processing of such dates easier.
      </p>
      <p>
        This strategy ensures that grepping for <code>;---</code>
        will always yield all the comments that require caution,
        as well as whom to talk to about each one.
      </p>
      <p>
        Only use <code>;---</code> on the first line of such a comment.
        Other lines should use spaces to align vertically.
        This way, grepping will also yield a count of the number of issues.
      </p>
      <p>
        You should insert a space after this comment prefix.
      </p>
      <p>
        You may use these with multiple-semicolon comments as well.
      </p>
      <p>
        Some people like to use words like <code>FIXME</code> or <code>TODO</code>.
        You may use these, but they must be preceded with <code>---</code>.
      </p>
      <p>
        Use TODO comments when the code is known to be incomplete
        and you want to indicate what work remains to be done.
      </p>
      <p>
        The comments begin with <code>TODO</code> in all capital letters,
        followed by your email address or other identifier in parentheses,
        followed by a colon, a space, and
        an explanation of what additional work is desirable or required.
        The user name included in the comment is that
        of a person who understands the deficiency.
        A TODO comment is not a commitment to fix the problem.
      </p>
      <p>
        When signing comments,
        you should use your username (for code within the company)
        or full email address (for code visible outside the company),
        not just initials.
        
      </p>
      <CODE_SNIPPET>
        ;;--- TODO(george@gmail.com): Refactor to provide a better API.
      </CODE_SNIPPET>
      <p>
        Be specific when indicating times or software releases
        in a TODO comment:
      </p>
      <CODE_SNIPPET>
        ;;--- TODO(brown): Remove this code after release 1.7 or before November, 2012.
      </CODE_SNIPPET>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Domain-Specific Languages">
    <SUMMARY>
      You should document DSLs and
      any terse program in a DSL.
    </SUMMARY>
    <BODY>
      <p>
        You should design your Domain Specific Language
        to be easy to read and understand by people familiar with the domain.
      </p>
      <p>
        You must properly document all your Domain Specific Language.
      </p>
      <p>
        Sometimes, your DSL is designed for terseness.
        In that case, it is important to document what each program does,
        if it's not painfully obvious from the context.
      </p>
      <p>
        Notably, when you use regular expressions
        (e.g. with the <code>CL-PPCRE</code> package),
        you MUST ALWAYS put in a comment
        (usually a two-semicolon comment on the previous line)
        explaining, at least basically, what the regular expression does,
        or what the purpose of using it is.
        The comment need not spell out every bit of the syntax, but
        it should be possible for someone to follow the logic of the code
        without actually parsing the regular expression.
      </p>
    </BODY>
  </STYLEPOINT>

</CATEGORY>

<CATEGORY title="Naming">
  <STYLEPOINT title="Symbol guidelines">
    <SUMMARY>
      You should use lower case.
      You should not abbreviate.
      You should follow punctuation conventions.
    </SUMMARY>
    <BODY>
      <p>
        Use lower case for all symbols.
        Consistently using lower case makes searching for symbol names easier
        and is more readable.
      </p>
      <p>
        Note that Common Lisp is case-converting,
        and that the <code>symbol-name</code> of your symbols
        will be upper case.
        Because of this case-converting,
        attempts to distinguish symbols by case are defeated,
        and only result in confusion.
        While it is possible to escape characters in symbols
        to force lower case,
        you should not use this capability
        unless this is somehow necessary
        to interoperate with third-party software.
      </p>
      <p>
        Place hyphens between all the words in a symbol.
        If you can't easily say an identifier out loud,
        it is probably badly named.
      </p>
      <p>
        You must not use <code>"/"</code> or <code>"."</code>
        instead of <code>"-"</code>
        unless you have a well-documented overarching reason to,
        and permission from other hackers who review your proposal.
      </p>
      <p>
        Generally, you should do not abbreviate words.
        You must avoid using abbreviations for words,
        unless it's a word that is used very frequently,
        in which case you must use
        the <em>same</em> abbreviation <em>consistently</em>.
        Abbreviations may also be used sparingly to avoid overly-long symbol names;
        it's easy to run into 100-column limit when there are very long names!
        You must especially avoid inconsistent abbreviations in exported names.
        For lexical variables of limited scope, abbreviations are fine.
      </p>
      <p>
      </p>
      <BAD_CODE_SNIPPET>
        ;; Bad
        (defvar *default-username* "Ann")
        (defvar *max-widget-cnt* 200)
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        ;; Better
        (defvar *default-user-name* "Ann")
        (defvar *maximum-widget-count* 200)
      </CODE_SNIPPET>
      <p>
        There are conventions in Common Lisp
        for the use of punctuation in symbols.
        You should not use punctuation in symbols outside these conventions.
      </p>
      <p>
        Unless the scope of a variable is very small,
        do not use overly short names like
        <code>i</code> and <code>zq</code>.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Denote intent, not content">
    <SUMMARY>
      Name your variables according to their intent,
      not their content.
    </SUMMARY>
    <BODY>
      <p>
        You should name a variable according
        to the high-level concept that it represents,
        not according to the low-level implementation details
        of how the concept is represented.
      </p>
      <p>
        Thus, you should avoid embedding
        data structure or aggregate type names,
        such as <code>list</code>, <code>array</code>, or
        <code>hash-table</code> inside variable names,
        unless you're writing a generic algorithm that applies to
        arbitrary lists, arrays, hash-tables, etc.
        In that case it's perfectly OK to name a variable
        <code>list</code> or <code>array</code>.
      </p>
      <p>
        Indeed, you should be introducing new abstract data types
        with <code>DEFCLASS</code> or <code>DEFTYPE</code>,
        whenever a new kind of intent appears for objects in your protocols.
        Functions that manipulate such objects generically may then
        use variables the name of which reflect that abstract type.
      </p>
      <p>
        For example, if a variable's value is always a row
        (or is either a row or <code>NIL</code>),
        it's good to call it <code>row</code> or <code>first-row</code>
        or something like that.
        It is alright is <code>row</code> has been
        <code>DEFTYPE</code>'d to <code>STRING</code> &#8212;
        precisely because you have abstracted the detail away,
        and the remaining salient point is that it is a row.
        You should not name the variable <code>STRING</code> in this context,
        except possibly in low-level functions that specifically manipulate
        the innards of rows to provide the suitable abstraction.
      </p>
      <p>
        Be consistent.
        If a variable is named <code>row</code> in one function,
        and its value is being passed to a second function,
        then call it <code>row</code> rather than, say, <code>value</code>
        (this was a real case).
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Global variables and constants">
    <SUMMARY>
      Name globals according to convention.
    </SUMMARY>
    <BODY>
      <p>
        The names of global constants should start and end
        with plus characters.
      </p>
      <p>
        Global variable names should start and end with asterisks
        (also known in this context as earmuffs).
      </p>
      <p>
        In some projects, parameters that are not meant
        to be usually modified or bound under normal circumstances
        (but may be during experimentation or exceptional situations)
        should start (but do not end) with a dollar sign.
        If such a convention exists within your project,
        you should follow it consistently.
        Otherwise, you should avoid naming variables like this.
      </p>
      <p>
        Common Lisp does not have global lexical variables,
        so a naming convention is used to ensure that globals,
        which are dynamically bound,
        never have names that overlap with local variables.
        It is possible to fake global lexical variables
        with a differently named global variable
        and a <code>DEFINE-SYMBOL-MACRO</code>.
        You should not use this trick,
        unless you first publish a library that abstracts it away.
      </p>
      <CODE_SNIPPET>
        (defconstant +hash-results+ #xbd49d10d10cbee50)

        (defvar *maximum-search-depth* 100)
      </CODE_SNIPPET>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Predicate names">
    <SUMMARY>
      Names of predicate functions end with a <code>"P"</code>.
    </SUMMARY>
    <BODY>
      <p>
        You should name boolean-valued functions with a trailing
        <code>"P"</code> or <code>"-P"</code>,
        to indicate they are predicates.
        Generally, you should use
        <code>"P"</code> when the rest of the function name is one word
        and <code>"-P"</code> when it is more than one word.
      </p>
      <p>
        A rationale for this convention is given in
        <a href="http://www.cs.cmu.edu/Groups/AI/html/cltl/clm/node69.html">the CLtL2 chapter on predicates</a>.
      </p>
      <p>
        For uniformity, you should follow the convention above,
        and not one of the alternatives below.
      </p>
      <p>
        An alternative rule used in some existing packages
        is to always use <code>"-P"</code>.
        Another alternative rule used in some existing packages
        is to always use <code>"?"</code>.
        When you develop such a package,
        you must be consistent with the rest of the package.
        When you start a new package,
        you should not use such an alternative rule
        without a very good documented reason.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Omit library prefixes">
    <SUMMARY>
      You should not include a library or package name
      as a prefix within the name of symbols.
    </SUMMARY>
    <BODY>
      <p>
        When naming a symbol (external or internal) in a package,
        you should not include the package name
        as a prefix within the name of the symbol.
        Naming a symbol this way makes it awkward to use
        from a client package accessing the symbol
        by qualifying it with a package prefix,
        where the package name then appears twice
        (once as a package prefix,
        another time as a prefix within the symbol name).
      </p>
      <BAD_CODE_SNIPPET>
        ;; Bad
        (in-package #:varint)
        (defun varint-length64 () ... )

        (in-package #:client-code)
        (defconst +padding+ (varint:varint-length64 +end-token+))
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        ;; Better
        (in-package #:varint)
        (defun length64 () ... )

        (in-package #:client-code)
        (defconst +padding+ (varint:length64 +end-token+))
      </CODE_SNIPPET>
      <p>
        An exception to the above rule would be to include a prefix
        for the names of variables that would otherwise be expected to clash
        with variables in packages that use the current one.
        For instance, <code>ASDF</code> exports a variable <code>*ASDF-VERBOSE*</code>
        that controls the verbosity of <code>ASDF</code> only,
        rather than of the entire Lisp program.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Packages">
    <SUMMARY>
      Use packages appropriately.
    </SUMMARY>
    <BODY>
      <p>
        Lisp packages are used to demarcate namespaces.
        Usually, each system has its own namespace.
        A package has a set of external symbols,
        which are intended to be used from outside the package,
        in order to allow other modules to use this module's facilities.
      </p>
      <p>
        The internal symbols of a package
        should never be referred to from other packages.
        That is, you should never have to use
        the double-colon <code>::</code> construct.
        (e.g. <code>QUAKE::HIDDEN-FUNCTION</code>).
        If you need to use double-colons to write real production code,
        something is wrong and needs to be fixed.
      </p>
      <p>
        As an exception,
        unit tests may use the internals of the package being tested.
        So when you refactor, watch out for
        internals used by the package's unit tests.
      </p>
      <p>
        The <code>::</code> construct is also useful for very temporary hacks,
        and at the REPL.
        But if the symbol really is part of
        the externally-visible definition of the package,
        export it.
      </p>
      <p>
        You may find that some internal symbols represent concepts
        you usually want to abstract away and hide under the hood,
        yet at times are necessary to expose for various extensions.
        For the former reason, you do not want to export them,
        yet for the latter reason, you have to export them.
        The solution is to have two different packages,
        one for your normal users to use, and
        another for the implementation and its extenders to use.
      </p>
      <p>
        Each package is one of two types:
      </p>
      <ul>
        <li>
          Intended to be included
          in the <code>:use</code> specification of other packages.
          If package <code>A</code> "uses" package <code>B</code>,
          then the external symbols of package <code>B</code>
          can be referenced from within package <code>A</code>
          without a package prefix.
          We mainly use this for low-level modules
          that provide widely-used facilities.
        </li>
        <li>
          Not intended to be "used".
          To reference a facility provided by package <code>B</code>,
          code in package <code>A</code> must use an explicit package prefix,
          e.g. <code>B:DO-THIS</code>.
        </li>
      </ul>
      <p>
        If you add a new package, it should always be of the second type,
        unless you have a special reason and get permission.
        Usually a package is designed to be one or the other,
        by virtue of the names of the functions.
        For example, if you have an abstraction called <code>FIFO</code>,
        and it were in a package of the first type
        you'd have functions named things like
        <code>FIFO-ADD-TO</code> and <code>FIFO-CLEAR-ALL</code>.
        If you used a package of the second type,
        you'd have names like <code>ADD-TO</code> and <code>CLEAR-ALL</code>,
        because the callers would be saying
        <code>FIFO:ADD-TO</code> and <code>FIFO:CLEAR-ALL</code>.
        (<code>FIFO:FIFO-CLEAR-ALL</code> is redundant and ugly.)
      </p>
      <p>
        Another good thing about packages is that
        your symbol names won't "collide" with the names of other packages,
        except the ones your packages "uses".
        So you have to stay away from symbols
        that are part of the Lisp implementation (since you always "use" that)
        and that are part of any other packages you "use",
        but otherwise you are free to make up your own names,
        even short ones, and not worry about some else
        having used the same name.
        You're isolated from each other.
      </p>
      <p>
        Your package must not shadow (and thus effectively redefine)
        symbols that are part of the Common Lisp language.
        There are certain exceptions,
        but they should be very well-justified and extremely rare:
      </p>
      <ul>
        <li>
          If you are explicitly replacing a Common Lisp symbol
          by a safer or more featureful version.
        </li>
        <li>
          If you are defining a package not meant to be "used",
          and have a good reason to export a symbol
          that clashes with Common Lisp,
          such as <code>log:error</code> and <code>log:warn</code>
          and so on.
        </li>
      </ul>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Language usage guidelines">
  <STYLEPOINT title="Mostly Functional Style">
    <SUMMARY>
      You should avoid side-effects when they are not necessary.
    </SUMMARY>
    <BODY>
      <p>
        Lisp is best used as a "mostly functional" language.
      </p>
      <p>
        Avoid modifying local variables, try rebinding instead.
      </p>
      <p>
        Avoid creating objects and the SETFing their slots.
        It's better to set the slots during initialization.
      </p>
      <p>
        Make classes as immutable as possible, that is, avoid giving slots
        setter functions if at all possible.
      </p>
      <p>
        Using a mostly functional style makes it much easier
        to write concurrent code that is thread-safe.
        It also makes it easier to test the code.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Recursion">
    <SUMMARY>
      You should favor iteration over recursion.
    </SUMMARY>
    <BODY>
      <p>
        Common Lisp systems are not required to implement
        function calls from tail positions without leaking stack space
        &#8212; which is known as proper tail calls (PTC),
        tail call elimination (TCE),
        or tail call optimization (TCO).
        This means that indefinite recursion through tail calls
        may quickly blow out the stack,
        which hampers functional programming.
        Still, most serious implementations (including SBCL and CCL)
        do implement proper tail calls, but with restrictions:
      </p>
      <ul>
        <li>
          The <code>(DECLARE (OPTIMIZE ...))</code> settings
          must favor <code>SPEED</code> enough and
          not favor <code>DEBUG</code> too much,
          for some compiler-dependent meanings of "enough" and "too much".
          (For instance, in SBCL, you should avoid <code>(SPEED 0)</code>
          and <code>(DEBUG 3)</code> to achieve proper tail calls.)
        </li>
        <li>
          There should not be dynamic bindings around the call
          (even though some Scheme compilers are able to properly treat
          such dynamic bindings, called parameters in Scheme parlance).
        </li>
      </ul>
      <p>
        For compatibility with all compilers and optimization settings,
        and to avoid stack overflow when debugging,
        you should prefer iteration or the built in mapping functions
        to relying on proper tail calls.
      </p>
      <p>
        If you do rely on proper tail calls,
        you must prominently document the fact,
        and take appropriate measures to ensure an appropriate compiler is used
        with appropriate optimization settings.
        For fully portable code, you may have to use trampolines instead.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Special variables">
    <SUMMARY>
      Use special variables sparingly.
    </SUMMARY>
    <BODY>
      <p>
        Using Lisp "special" (dynamically bound) variables
        as implicit arguments to functions should be used sparingly,
        and only in cases where it won't surprise the person reading the code,
        and where it offers significant benefits.
      </p>
      <p>
        Indeed, each special variable constitutes state.
        Developers have to mentally track the state of all relevant variables
        when trying to understand what the code does and how it does it;
        tests have to be written and run with all relevant combinations;
        to isolate some activity, care has to be taken to locally bind
        all relevant variables, including those of indirectly used modules.
        They can hide precious information from being printed in a backtrace.
        Not only is there overhead associated to each new variable,
        but interactions between variables
        can make the code exponentially more complex
        as the number of such variables increases.
        The benefits have to match the costs.
      </p>
      <p>
        Note though that a Lisp special variable is not a global variable
        in the sense of a global variable in, say, BASIC or C.
        As special variables can be dynamically bound to a local value,
        they are much more powerful than
        global value cells where all users necessarily interfere with each other.
      </p>
      <p>
        Good candidates for such special variables
        are items for which "the current" can be naturally used as prefix,
        such as "the current database connection" or
        "the current business data source".
        They are singletons as far as the rest of the code is concerned,
        and often passing them as an explicit argument
        does not add anything to the readability or maintainability
        of the source code in question.
      </p>
      <p>
        They can make it easier to write code that can be refactored.
        If you have a request processing chain,
        with a number of layers that all operate upon a "current" request,
        passing the request object explicitly to every function
        requires that every function in the chain have a request argument.
        Factoring out code into new functions often requires
        that these functions also have this argument,
        which clutters the code with boilerplate.
      </p>
      <p>
        You should treat special variables
        as though they are per-thread variables.
        By default, you should leave a special variable
        with no top-level binding at all,
        and each thread of control
        that needs the variable should bind it explicitly.
        This will mean that any incorrect use of the variable
        will result in an "unbound variable" error, and
        each thread will see its own value for the variable.
        Variables with a default global value should usually be
        locally bound at thread creation time.
        You should use suitable infrastructure
        to automate the appropriate declaration of such variables.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Assignment">
    <SUMMARY>
      Be consistent in assignment forms.
    </SUMMARY>
    <BODY>
      <p>
        There are several styles for dealing with assignment and side-effects;
        whichever a given package is using,
        keep using the same consistently when hacking said package.
        Pick a style that makes sense when starting a new package.
      </p>
      <p>
        Regarding multiple assignment in a same form, there are two schools:
        the first style groups as many assignments as possible into a single
        <code>SETF</code> or <code>PSETF</code> form
        thus minimizing the number of forms with side-effects;
        the second style splits assignments into as many individual
        <code>SETF</code> (or <code>SETQ</code>, see below) forms as possible,
        to maximize the chances of locating forms that modify a kind of place
        by grepping for <code>(setf (foo ...</code>.
        A grep pattern must actually contain as many place-modifying forms
        as you may use in your programs, which may make this rationale either
        convincing or moot depending on the rest of the style of your code.
        You should follow the convention used in the package you are hacking.
        We recommend the first convention for new packages.
      </p>
      <p>
        Regarding <code>SETF</code> and <code>SETQ</code>,
        there are two schools:
        this first regards <code>SETQ</code>
        as an archaic implementation detail,
        and avoids it entirely in favor of <code>SETF</code>;
        the second regards <code>SETF</code>
        as an additional layer of complexity,
        and avoids it in favor of <code>SETQ</code> whenever possible
        (i.e. whenever the assigned place is a variable or symbol-macro).
        You should follow the convention used in the package you are hacking.
        We recommend the first convention for new packages.
      </p>
      <p>
        In the spirit of a mostly pure functional style,
        which makes testing and maintenance easier,
        we invite you to consider how to do things with the fewest assignments required.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Assertions and Conditions">
    <SUMMARY>
      You must make proper usage of assertions and conditions.
    </SUMMARY>
    <BODY>
      <ul>
        <li>
          <code>ASSERT</code> should be used ONLY to detect internal bugs.
          Code should <code>ASSERT</code> invariants whose failure indicates
          that the software is itself broken.
          Incorrect input should be handled properly at runtime,
          and must not cause an assertion violation.
          The audience for an <code>ASSERT</code> failure is a developer.
          Do not use the data-form and argument-form in <code>ASSERT</code>
          to specify a condition to signal.
          It's fine to use them to print out a message for debugging purposes
          (and since it's only for debugging, there's no issue of
          internationalization).
        </li>
        <li>
          <code>CHECK-TYPE</code>,
          <code>ETYPECASE</code> are also forms of assertion.
          When one of these fails, that's a detected bug.
          You should prefer to use <code>CHECK-TYPE</code>
          over (DECLARE (TYPE ...))
          for the inputs of functions.
        </li>
        <li>
          Your code should use assertions and type checks liberally.
          The sooner a bug is discovered, the better!
          Only code in the critical path for performance
          and internal helpers should eschew
          explicit assertions and type checks.
        </li>
        <li>
          Invalid input, such as files that are read
          but do not conform to the expected format,
          should not be treated as assertion violations.
          Always check to make sure that input is valid,
          and take appropriate action if it is not,
          such as signalling a real error.
        </li>
        <li>
          <code>ERROR</code> should be used
          to detect problems with user data, requests, permissions, etc.,
          or to report "unusual outcomes" to the caller.
        </li>
        <li>
          <code>ERROR</code> should always be called
          with an explicit condition type;
          it should never simply be called with a string.
          This enables internationalization.
        </li>
        <li>
          Functions that report unusual outcomes
          by signaling a condition should say so explicitly in their contracts
          (their textual descriptions, in documentation and docstrings etc.).
          When a function signals a condition
          that is not specified by its contract, that's a bug.
          The contract should specify the condition class(es) clearly.
          The function may then signal any condition
          that is a type-of any of those conditions.
          That is, signaling instances of subclasses
          of the documented condition classes is fine.
        </li>
        <li>
          Complex bug-checks may need to use <code>ERROR</code>
          instead of <code>ASSERT</code>.
          
        </li>
        <li>
          When writing a server, you must not call <code>WARN</code>.
          Instead, you should use the appropriate logging framework.
          
        </li>
        <li>
          Code must not call <code>SIGNAL</code>.
          Instead, use <code>ERROR</code> or <code>ASSERT</code>.
        </li>
        <li>
          Code should not use <code>THROW</code> and <code>CATCH</code>;
          instead use the <code>restart</code> facility.
        </li>
        <li>
          Code should not generically handle all conditions,
          e.g. type <code>T</code>, or use <code>IGNORE-ERRORS</code>.
          Instead, let unknown conditions propagate to
          the standard ultimate handler for processing.
          
        </li>
        <li>
          There are a few places where handling all conditions is appropriate,
          but they are rare.
          The problem is that handling all conditions can mask program bugs.
          If you <em>do</em> need to handle "all conditions",
          you MUST handle only <code>ERROR</code>, <em>not</em> <code>T</code>
          and not <code>SERIOUS-CONDITION</code>.
          (This is notably because CCL's process shutdown
          depends on being able to signal <code>process-reset</code>
          and have it handled by CCL's handler,
          so we must not interpose our own handler.)
        </li>
        <li>
          <code>(error (make-condition 'foo-error ...))</code>
          is equivalent to <code>(error 'foo-error ...)</code> &#8212;
          code must use the shorter form.
        </li>
        <li>
          Code should not signal conditions from inside the cleanup form of
          <code>UNWIND-PROTECT</code>
          (unless they are always handled inside the cleanup form),
          or otherwise do non-local exits from cleanup handlers
          outside of the handler e.g. <code>INVOKE-RESTART</code>.
        </li>
        <li>
          Do not clean up by resignaling.
          If you do that, and the condition is not handled,
          the stack trace will halt at the point of the resignal,
          hiding the rest.
          And the rest is the part we really care about!
          <BAD_CODE_SNIPPET>
            ;; Bad
            (handler-case
              (catch 'ticket-at
                (etd-process-blocks))
              (error (c)
                (reset-parser-values)
                  (error c)))
          </BAD_CODE_SNIPPET>
          <CODE_SNIPPET>
            ;; Better
            (unwind-protect
              (catch 'ticket-at
                (etd-process-blocks))
              (reset-parser-values))
          </CODE_SNIPPET>
        </li>
      </ul>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Type Checking">
    <SUMMARY>
      If you know the type of something, you should make it explicit
      in order to enable compile-time and run-time sanity-checking.
    </SUMMARY>
    <BODY>
      
      <p>
        If your function is using a special variable as an implicit argument,
        it's good to put in a <code>CHECK-TYPE</code> for the special variable,
        for two reasons:
        to clue in the person reading the code
        that this variable is being used implicitly as an argument,
        and also to help detect bugs.
      </p>
      
      <p>
        Using <code>(declare (type ...))</code>
        is the least-desirable mechanism to use
        because, as Scott McKay puts it:
      </p>
      <blockquote>
        <p>
          The fact is, <code>(declare (type ...))</code> does different things
          depending on the compiler settings of speed, safety, etc.
          In some compilers, when speed is greater than safety,
          <code>(declare (type ...))</code> will tell the compiler
          "please assume that these variables have these types"
          <em>without</em> generating any type-checks.
          That is, if some variable has the value <code>1432</code> in it,
          and you declare it to be of type <code>string</code>,
          the compiler might just go ahead and use it as though it's a string.
        </p>
        <p>
          Moral: don't use <code>(declare (type ...))</code>
          to declare the contract of any API functions,
          it's not the right thing.
          Sure, use it for "helper" functions, but not API functions.
        </p>
      </blockquote>
      <p>
        You should, of course, use appropriate declarations
        in internal low-level functions
        where these declarations are used for optimization.
        When you do, however, see our recommendations for
        <a href="#Unsafe_Operations">Unsafe Operations</a>.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="CLOS">
    <SUMMARY>
      Use CLOS appropriately.
    </SUMMARY>
    <BODY>
      <p>
        When a generic function is intended to be called from other
        modules (other parts of the code), there should be an
        explicit <code>DEFGENERIC</code> form,
        with a <code>:DOCUMENTATION</code> string
        explaining the generic contract of the function
        (as opposed to its behavior for some specific class).
        It's generally good to do explicit <code>DEFGENERIC</code> forms,
        but for module entry points it is mandatory.
      </p>
      <p>
        When the argument list of a generic function includes
        <code>&amp;KEY</code>,
        the <code>DEFGENERIC</code> should always explicitly list
        all of the keyword arguments that are acceptable,
        and explain what they mean.
        (Common Lisp does not require this, but it is good form,
        and it may avoid spurious warnings on SBCL.)
      </p>
      <p>
        You should avoid <code>SLOT-VALUE</code> and <code>WITH-SLOTS</code>,
        unless you absolutely intend to circumvent
        any sort of method combination that might be in effect for the slot.
        Rare exceptions include <code>INITIALIZE-INSTANCE</code>
        and <code>PRINT-OBJECT</code> methods and
        accessing normally hidden slots in the low-level implementation of
        methods that provide user-visible abstractions.
        Otherwise, you should use accessors,
        <code>WITH-ACCESSORS</code>
      </p>
      
      <p>
        Accessor names generally follow a convention of
        <code>&lt;protocol-name&gt;-&lt;slot-name&gt;</code>,
        where a "protocol" in this case loosely indicates
        a set of functions with well-defined behavior.
      </p>
      <p>
        No implication of a formal "protocol" concept is necessarily intended,
        much less first-class "protocol" objects.
        However, there may indeed be an abstract CLOS class
        or an
        <a href="http://common-lisp.net/~frideau/lil-ilc2012/lil-ilc2012.html">Interface-Passing Style</a> interface
        that embodies the protocol.
        Further (sub)classes or (sub)interfaces may then implement
        all or part of a protocol by defining
        some methods for (generic) functions in the protocol,
        including readers and writers.
      </p>
      <p>
        For example, if there were a notional protocol called
        is <code>pnr</code> with accessors <code>pnr-segments</code>
        and <code>pnr-passengers</code>, then
        the classes <code>air-pnr</code>, <code>hotel-pnr</code> and
        <code>car-pnr</code> could each reasonably implement
        methods for <code>pnr-segments</code> and <code>pnr-passengers</code>
        as accessors.
      </p>
      <p>
        By default, an abstract base class name is used
        as the notional protocol name, so accessor names default
        to <code>&lt;class-name&gt;-&lt;slot-name&gt;</code>;
        while such names are thus quite prevalent,
        this form is neither required nor even preferred.
        In general, it contributes to "symbol bloat",
        and in many cases has led to a proliferation of "trampoline" methods.
      </p>
      <p>
        Accessors named <code>&lt;slot-name&gt;-of</code> should not be used.
      </p>
      <p>
        Explicit <code>DEFGENERIC</code> forms should be used when there are
        (or it is anticipated that there will be)
        more than one <code>DEFMETHOD</code> for that generic function.
        The reason is that the documentation for the generic function
        explains the abstract contract for the function,
        as opposed to explaining what an individual method does for
        some specific class(es).
      </p>
      <p>
        You must not use generic functions where there is no notional protocol.
        To put it more concretely,
        if you have more than one generic function that specializes its Nth argument,
        the specializing classes should all be descendants of a single class.
        Generic functions must not be used for "overloading",
        i.e. simply to use the same name for two entirely unrelated types.
      </p>
      <p>
        More precisely, it's not really
        whether they descend from a common superclass,
        but whether they obey the same "protocol".
        That is, the two classes should handle the same set of generic functions,
        as if there were an explicit <code>DEFGENERIC</code> for each method.
      </p>
      <p>
        Here's another way to put it.
        Suppose you have two classes, A and B, and a generic function F.
        There are two methods for F,
        which dispatch on an argument being of types A and B.
        Is it plausible that there might be a function call somewhere
        in the program that calls F,
        in which the argument might sometimes, at runtime,
        be of class A and other times be of class B?
        If not, you probably are overloading and
        should not be using a single generic function.
      </p>
      <p>
        We allow one exception to this rule:
        it's OK to do overloading
        if the corresponding argument "means" the same thing.
        Typically one overloading allows an X object,
        and the other allows the name of an X object,
        which might be a symbol or something.
      </p>
      <p>
        You must not use MOP "intercessory" operations at runtime.
        You should not use MOP "intercessory" operations at compile-time.
        At runtime, they are at worst a danger, at best a performance issue.
        At compile-time, it is usually cleaner that
        macros should set things up the right way in one pass
        than have to require a second pass of fixups through intercession;
        but sometimes, fixups are necessary to resolve forward references,
        and intercession is allowed then.
        MOP intercession is a great tool for interactive development,
        and you may enjoy it while developping and debugging;
        but you should not use it in normal applications.
      </p>
      <p>
        If a class definition creates a method
        as a <code>:READER</code>, <code>:WRITER</code>,
        or <code>:ACCESSOR</code>,
        do not redefine that method.
        It's OK to add <code>:BEFORE</code>, <code>:AFTER</code>,
        and <code>:AROUND</code> methods,
        but don't override the primary method.
      </p>
      <p>
        In methods with keyword arguments,
        you must always use <code>&amp;KEY</code>,
        even if the method does not care about the values of any keys,
        and you should never use <code>&amp;ALLOW-OTHER-KEYS</code>.
        As long as a keyword is accepted by any method of a generic function,
        it's OK to use it in the generic function,
        even if the other methods of the same generic function
        don't mention it explicitly.
        This is particularly important
        for <code>INITIALIZE-INSTANCE</code> methods,
        since if you did use <code>&amp;ALLOW-OTHER-KEYS</code>,
        it would disable error checking for misspelled or wrong keywords
        in <code>MAKE-INSTANCE</code> calls!
      </p>
      
      <p>
        A typical <code>PRINT-OBJECT</code> method might look like this:
      </p>
      <CODE_SNIPPET>
        (defmethod print-object ((p person) stream)
          (print-unprintable-object (p stream :type t :identity t)
            (with-slots (first-name last-name) p
              (safe-format stream "~a ~a" first-name last-name))))
      </CODE_SNIPPET>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Meta-language guidelines">
  <STYLEPOINT title="Macros">
    <SUMMARY>
      Use macros when appropriate, which is often.
      Define macros when appropriate, which is seldom.
    </SUMMARY>
    <BODY>
      <p>
        Macros bring syntactic abstraction, which is a wonderful thing.
        It helps make your code clearer, by describing your intent
        without getting bogged in implementation details
        (indeed abstracting those details away).
        It helps make your code more concise and more readable,
        by eliminating both redundancy and irrelevant details.
        But it comes at a cost to the reader,
        which is learning a new syntactic concept for each macro.
        And so it should not be abused.
      </p>
      <p>
        The general conclusion is that there shouldn't be
        any recognizable <em>design pattern</em>
        in a good Common Lisp program.
        The one and only pattern is: <em>use the language</em>,
        which includes defining and using syntactic abstractions.
      </p>
      <p>
        Existing macros must be used
        whenever they make code clearer
        by conveying intent in a more concise way,
        which is often.
        When a macro is available in your project
        that expresses the concept you're using,
        you must not write the expansion rather than use the macro.
      </p>
      <p>
        New macros should be defined as appropriate,
        which should be seldom,
        for common macros have already been provided
        by the language and its various libraries,
        and your program typically only needs few new ones
        relative to its size.
      </p>
      <p>
        You should follow the OAOOM rule of thumb
        for deciding when to create a new abstraction,
        whether syntactic or not:
        if a particular pattern is used more than twice,
        it should probably be abstracted away.
        A more refined rule to decide when to use abstraction
        should take into account
        the benefit in term of number of uses and gain at each use,
        to the costs in term of having to get used to reading the code.
        For syntactic abstractions, costs and benefits to the reader
        is usually more important than costs and benefits to the writer,
        because good code is usually written once
        and read many times by many people
        (including the same programmer
        who has to maintain the program after having forgotten it).
        Yet the cost to the writer of the macro
        should also be taken into account;
        however, in doing so it should rather be compared
        to the cost of the programmer writing other code instead
        that may have higher benefits.
      </p>
      <p>
        Using Lisp macros properly requires taste.
        Avoid writing complicated macros
        unless the benefit clearly outweighs the cost.
        It takes more effort for your fellow developers to learn your macro,
        so you should only use a macro if the gain in expressiveness
        is big enough to justify that cost.
        As usual, feel free to consult your colleagues if you're not sure,
        since without a lot of Lisp experience,
        it can be hard to make this judgment.
      </p>
      <p>
        You must never use a macro where a function will do.
        That is, if the semantics of what you are writing
        conforms to the semantics of a function,
        then you must write it as a function rather than a macro.
      </p>
      <p>
        You must not transform a function into a macro for performance reasons.
        If profiling shows that you have a performance problem
        with a specific function <code>FOO</code>,
        document the need and profiling-results appropriately,
        and
        <code>(declaim (inline foo))</code>.
      </p>
      
      <p>
        You can also use a compiler-macro
        as a way to speed up function execution
        by specifying a source-to-source transformation.
        Beware that it interferes with tracing the optimized function.
      </p>
      <p>
        When you write a macro-defining macro
        (a macro that generates macros),
        document and comment it particularly clearly,
        since these are hard for the uninitiated to understand.
      </p>
      <p>
        You must not install new reader macros
        without a consensus among the developers of your system.
        Reader macros must not leak out of the system that uses them
        to clients of that system or other systems used in the same project.
        You must software such as
        <code>cl-syntax</code> or <code>named-readtables</code>
        to control how reader macros are used.
        This clients who desire it may use the same reader macros as you do.
        In any case, your system must be usable
        even to clients who do not use these reader macros.
      </p>
      <p>
        If your macro has a parameter that is a Lisp form
        that will be evaluated when the expanded code is run,
        you should name the parameter with the suffix <code>-form</code>.
        This convention helps make it clearer to the macro's user
        which parameters are Lisp forms to be evaluated, and which are not.
      </p>
      <p>
        You should follow the so-called <code>CALL-WITH</code> style when it applies.
        This style is explained at length in
        <a href="http://random-state.net/log/3390120648.html">http://random-state.net/log/3390120648.html</a>.
        The general principle is that the macro is strictly limited to processing the syntax,
        and as much of the semantics as possible is kept in normal functions.
        Therefore, a macro <code>WITH-<em>FOO</em></code> is often limited to
        generating a call to an auxiliary function
        <code>CALL-WITH-<em>FOO</em></code>
        with arguments deduced from the macro arguments.
        Macro <code>&amp;body</code> arguments are typically
        wrapped into a lambda expression of which they become the body,
        which is passed as one of the arguments of the auxiliary function.
      </p>
      <p>
        The separation of syntactic and semantic concerns
        is a general principle of style that applies
        beyond the case of <code>WITH-</code> macros.
        Its advantages are many.
        By keeping semantics outside the macro,
        the macro is made simpler, easier to get right, and less subject to change,
        which makes it easier to develop and maintain.
        The semantics is written in a simpler language &#8212; one without staging &#8212;
        which also makes it easier to develop and maintain.
        It becomes possible to debug and update the semantic function
        without having to recompile all clients of the macro.
        The semantic function appears in the stack trace
        which also helps debug client functions.
        The macro expansion is made shorter and
        each expansion shares more code with other expansions,
        which reduces memory pressure which in turn usually makes things faster.
        It also makes sense to write the semantic functions first,
        and write the macros last as syntactic sugar on top.
        You should use this style unless the macro is used
        in tight loops where performance matters;
        and even then, see our rules regarding optimization.
      </p>
      <p>
        Any functions (closures) created by the macro should be named,
        which can be done using <code>FLET</code>.
        
        This also allows you to declare the function to be of dynamic extent
        (if it is &#8212; and often it is; yet see below regarding
        <a href="#DYNAMIC-EXTENT">DYNAMIC-EXTENT</a>).
      </p>
      <p>
        If a macro call contains a form,
        and the macro expansion includes more than one copy of that form,
        the form can be evaluated more than once,
        and code it contains macro-expanded and compiled more than once.
        If someone uses the macro and calls it
        with a form that has side effects or that takes a long time to compute,
        the behavior will be undesirable
        (unless you're intentionally writing
        a control structure such as a loop).
        A convenient way to avoid this problem
        is to evaluate the form only once,
        and bind a (generated) variable to the result.
        There is a very useful macro called <code>ALEXANDRIA:ONCE-ONLY</code>
        that generates code to do this.
        See also <code>ALEXANDRIA:WITH-GENSYMS</code>,
        to make some temporary variables in the generated code.
        Note that if you follow our <code>CALL-WITH</code> style,
        you typically expand the code only once, as either
        an argument to the auxiliary function, or
        the body of a lambda passed as argument to it;
        you therefore avoid the above complexity.
      </p>
      <p>
        When you write a macro with a body,
        such as a <code>WITH-xxx</code> macro,
        even if there aren't any parameters,
        you should leave space for them anyway.
        For example, if you invent <code>WITH-LIGHTS-ON</code>,
        do not make the call to it look like
        <code>(defmacro with-lights-on (&amp;body b) ...)</code>.
        Instead, do <code>(defmacro with-lights-on (() &amp;body b) ...)</code>.
        That way, if parameters are needed in the future,
        you can add them without necessarily having to change
        all the uses of the macro.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="EVAL-WHEN">
    <SUMMARY>
      When using <code>EVAL-WHEN</code>, you should almost always use all of
      <code>(:compile-toplevel :load-toplevel :execute)</code>.
    </SUMMARY>
    <BODY>
      <p>
        Lisp evaluation happens at several times,
        some of them interleaved.
        Be aware of them when writing macros.
        <a href="http://fare.livejournal.com/146698.html">EVAL-WHEN considered harmful to your mental health</a>.
      </p>
      <p>
        In summary of the article linked above,
        unless you're doing truly advanced macrology,
        the only valid combination in an <code>EVAL-WHEN</code>
        is to include all of
        <code>(eval-when (:compile-toplevel :load-toplevel :execute) ...)</code>
      </p>
      <p>
        You must use
        <code>(eval-when (:compile-toplevel :load-toplevel :execute) ...)</code>
        whenever you define functions, types, classes, constants, variables, etc.,
        that are going to be used in macros.
      </p>
      <p>
        It is usually an error to omit the <code>:execute</code>,
        because it prevents <code>LOAD</code>ing the source rather than the fasl.
        It is usually an error to omit the <code>:load-toplevel</code>
        (except to modify e.g. readtables and compile-time settings),
        because it prevents <code>LOAD</code>ing future files
        or interactively compiling code
        that depends on the effects that happen at compile-time,
        unless the current file was <code>COMPILE-FILE</code>d
        within the same Lisp session.
      </p>
      <p>
        Regarding variables, note that because macros may or may not
        be expanded in the same process that runs the expanded code,
        you must not depend on compile-time and runtime effects
        being either visible or invisible at the other time.
        There are still valid uses of variables in macros:
      </p>
      <ul>
        <li>
          Some variables may hold dictionaries
          for some new kind of definition and other meta-data.
          If such meta-data is to be visible at runtime and/or in other files,
          you must make sure that the macro expands into code that
          will register the definitions to those meta-data structures
          at load-time,
          in addition to effecting the registration at compile-time.
          Typically, your top-level definitions expand
          to code that does the registration.
          if your code doesn't expand at the top-level,
          you can sometimes use <code>LOAD-TIME-VALUE</code> for good effect.
          In extreme cases, you may have to use
          <code>ASDF-FINALIZERS:EVAL-AT-TOPLEVEL</code>.
        </li>
        <li>
          Some variables may hold temporary data
          that is only used at compile-time in the same file,
          and can be cleaned up at the end of the file's compilation.
          Predefined such variables would include <code>*readtable*</code>
          or compiler-internal variables holding
          the current optimization settings.
          You can often manage existing and new such variables using
          the <code>:AROUND-COMPILE</code> hooks of <code>ASDF</code>.
        </li>
      </ul>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Read-time evaluation">
    <SUMMARY>
      You should use <code>#.</code> sparingly,
      and you must avoid read-time side-effects.
    </SUMMARY>
    <BODY>
      <p>
        The <code>#.</code> standard read-macro
        will read one object, evaluate the object, and
        have the reader return the resulting value.
      </p>
      <p>
        You must not use it where other idioms will do, such as
        using <code>EVAL-WHEN</code> to evaluate side-effects at compile-time,
        using a regular macro to return an expression computed at compile-time,
        using <code>LOAD-TIME-VALUE</code> to compute it at load-time.
      </p>
      <p>
        Read-time evaluation is often used as a quick way
        to get something evaluated at compile time
        (actually "read time" but it amounts to the same thing).
        If you use this, the evaluation MUST NOT have any side effects
        and MUST NOT depend on any variable global state.
        The <code>#.</code> should be treated as a way
        to force "constant-folding"
        that a sufficiently-clever compiler
        could have figure out all by itself,
        when the compiler isn't sufficiently-clever
        and the difference matters.
      </p>
      <p>
        Another use of <code>#.</code> is to expand the equivalent of macros
        in places that are neither expressions nor (quasi)quotations,
        such as lambda-lists. However, if you find yourself using it a lot,
        it might be time to instead define macros to replace your consumers
        of lambda-lists with something that recognizes an extension.
      </p>
      <p>
        Whenever you are going to use <code>#.</code>,
        you should consider using <code>DEFCONSTANT</code> and its variants,
        possibly in an <code>EVAL-WHEN</code>,
        to give the value a name explaining what it means.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="EVAL">
    <SUMMARY>
      You must not use <code>EVAL</code> at runtime.
    </SUMMARY>
    <BODY>
      <p>
        Places where it is actually appropriate to use <code>EVAL</code>
        are so few and far between that you must consult with your reviewers;
        it's easily misused.
      </p>
      <p>
        If your code manipulates symbols at runtime
        and needs to get the value of a symbol,
        use <code>SYMBOL-VALUE</code>, not <code>EVAL</code>.
      </p>
      <p>
        Often, what you really need is to write a macro,
        not to use <code>EVAL</code>.
      </p>
      <p>
        You may be tempted to use <code>EVAL</code> as a shortcut
        to evaluating expressions in a safe subset of the language.
        But it often requires more scrutiny to properly check and sanitize
        all possible inputs to such use of <code>EVAL</code>
        than to build a special-purpose evaluator.
        You must not use <code>EVAL</code> in this way at runtime.
      </p>
      <p>
        Places where it is OK to use <code>EVAL</code> are:
      </p>
      <ul>
        <li>
          Implementing an interactive development loop.
        </li>
        <li>
          The build infrastructure.
        </li>
        <li>
          Backdoors that are part of testing frameworks.
          (You MUST NOT have such backdoors in production code.)
        </li>
        <li>
          Macros that fold constants at compile-time.
        </li>
        <li>
          Macros that register definitions to meta-data structures;
          the registration form is sometimes evaluated at compile-time
          as well as included in the macro-expansion,
          so it is immediately available to other macros.
        </li>
      </ul>
      <p>
        Note that in the latter case,
        if the macro isn't going to be used at the top-level,
        it might not be possible to make these definitions available
        as part of the expansion.
        The same phenomenon may happen in a <code>DEFTYPE</code> expansion,
        or in helper functions used by macros.
        In these cases, you may actually have to use
        <code>ASDF-FINALIZERS:EVAL-AT-TOPLEVEL</code> in your macro.
        It will not only <code>EVAL</code> your definitions
        at macro-expansion time for immediate availability,
        it will also save the form aside, for inclusion in a
        <code>(ASDF-FINALIZERS:FINAL-FORMS)</code>
        that you need to include at the end of the file being compiled
        (or before the form is needed).
        This way, the side-effects are present when loading the fasl
        without having compiled it as well as while compiling it;
        in either case, the form is made available at load-time.
        <code>ASDF-FINALIZERS</code> ensures that the form is present,
        by throwing an error if you omit it.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="INTERN and UNINTERN">
    <SUMMARY>
      You must not use <code>INTERN</code> or <code>UNINTERN</code> at runtime.
    </SUMMARY>
    <BODY>
      <p>
        You must not use <code>INTERN</code> at runtime.
        Not only does it cons,
        it either creates a permanent symbol that won't be collected
        or gives access to internal symbols.
        This creates opportunities for memory leaks, denial of service attacks,
        unauthorized access to internals, clashes with other symbols.
      </p>
      <p>
        You must not <code>INTERN</code> a string
        just to compare it to a keyword;
        use <code>STRING=</code> or <code>STRING-EQUAL</code>.
      </p>
      <BAD_CODE_SNIPPET>
        (member (intern str :keyword) $keys) ; Bad
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        (member str $keys :test #'string-equal) ; Better
      </CODE_SNIPPET>
      <p>
        You must not use <code>UNINTERN</code> at runtime.
        It can break code that relies on dynamic binding.
        It makes things harder to debug.
        You must not dynamically intern any new symbol,
        and therefore you need not dynamically unintern anything.
      </p>
      <p>
        You may of course use <code>INTERN</code> at compile-time,
        in the implementation of some macros.
        Even so, it is usually more appropriate
        to use abstractions on top of it, such as
        <code>ALEXANDRIA:SYMBOLICATE</code> or
        <code>ALEXANDRIA:FORMAT-SYMBOL</code>
        to create the symbols you need.
      </p>
      
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Data Representation">
  <STYLEPOINT title="NIL: empty-list, false and I Don't Know">
    <SUMMARY>
      Appropriately use or avoid using <code>NIL</code>.
    </SUMMARY>
    <BODY>
      <p>
        <code>NIL</code> can have several different interpretations:
      </p>
      <ul>
        <li>
          "False."
          In this case, use <code>NIL</code>.
          You should test for false <code>NIL</code>
          using the operator <code>NOT</code> or
          using the predicate function <code>NULL</code>.
        </li>
        <li>
          "Empty-list."
          In this case, use <code>'()</code>.
          (Be careful about quoting the empty-list when calling macros.)
          You should use <code>ENDP</code> to test for the empty list
          when the argument is known to be a proper list,
          or with <code>NULL</code> otherwise.
        </li>
        <li>
          A statement about some value being unspecified.
          In this case, you may use <code>NIL</code>
          if there is no risk of ambiguity anywhere in your code;
          otherwise you should use an explicit, descriptive symbol.
        </li>
        <li>
          A statement about some value being known not to exist.
          In this case, you should use an explicit, descriptive symbol
          instead of <code>NIL</code>.
        </li>
      </ul>
      <p>
        You must not introduce ambiguity in your data representations
        that will cause headaches for whoever has to debug code.
        If there is any risk of ambiguity,
        you should use an explicit, descriptive symbol or keyword
        for each case,
        instead of using <code>NIL</code> for either.
        If you do use <code>NIL</code>,
        you must make sure that the distinction is well documented.
      </p>
      <p>
        In many contexts,
        instead of representing "I don't know" as a particular value,
        you should instead use multiple values,
        one for the value that is known if any,
        and one to denote whether the value was known or found.
      </p>
      
      <p>
        When working with database classes, keep in mind that
        <code>NIL</code> need not always map to <code>'NULL'</code>
        (and vice-versa)!
        The needs of the database may differ from the needs of the Lisp.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Do not abuse lists">
    <SUMMARY>
      You must select proper data representation.
      You must not abuse the <code>LIST</code> data structure.
    </SUMMARY>
    <BODY>
      <p>
        Even though back in 1958, LISP was short for "LISt Processing",
        its successor Common Lisp has been a modern programming language
        with modern data structures since the 1980s.
        You must use the proper data structures in your programs.
      </p>
      <p>
        You must not abuse the builtin (single-linked) <code>LIST</code>
        data structure where it is not appropriate,
        even though Common Lisp makes it especially easy to use it.
      </p>
      <p>
        You must only use lists
        when their performance characteristics
        is appropriate for the algorithm at hand:
        sequential iteration over the entire contents of the list.
      </p>
      <p>
        An exception where it is appropriate to use lists
        is when it is known in advance
        that the size of the list will remain very short
        (say, less than 16 elements).
      </p>
      <p>
        List data structures are often (but not always)
        appropriate for macros and functions used by macros at compile-time:
        indeed, not only is source code passed as lists in Common Lisp,
        but the macro-expansion and compilation processes
        will typically walk over the entire source code, sequentially, once.
        (Note that advanced macro systems don't directly use lists, but instead
        use abstract syntax objects that track source code location and scope;
        however there is no such advanced macro system
        in Common Lisp at this time.)
      </p>
      <p>
        Another exception where it is appropriate to use lists is
        for introducing literal constants
        that will be transformed into more appropriate data structures
        at compile-time or load-time.
        It is a good to have a function with a relatively short name
        to build your program's data structures from such literals.
      </p>
      <p>
        In the many cases when lists are not the appropriate data structure,
        various libraries such as
        <a href="http://cliki.net/cl-containers">cl-containers</a> or
        <a href="http://cliki.net/lisp-interface-library">lisp-interface-library</a>
        provide plenty of different data structures
        that should fulfill all the basic needs of your programs.
        If the existing libraries are not satisfactory, see above about
        <a href="#Using_Libraries">Using Libraries</a> and
        <a href="#Open-Sourcing_Code">Open-Sourcing Code</a>.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Lists vs. structures vs. multiple values">
    <SUMMARY>
      You should use the appropriate representation for product types.
    </SUMMARY>
    <BODY>
      <p>
        You should avoid using a list as anything
        besides a container of elements of like type.
        You must not use a list as method of passing
        multiple separate values of different types
        in and out of function calls.
        Sometimes it is convenient to use a list
        as a little ad hoc structure,
        i.e. "the first element of the list is a FOO, and the second is a BAR",
        but this should be used minimally
        since it gets harder to remember the little convention.
        You must only use a list that way
        when destructuring the list of arguments from a function,
        or creating a list of arguments
        to which to <code>APPLY</code> a function.
      </p>
      <p>
        The proper way to pass around an object
        comprising several values of heterogeneous types
        is to use a structure as defined by <code>DEFSTRUCT</code>
        or <code>DEFCLASS</code>.
      </p>
      <p>
        You should use multiple values only
        when function returns a small number of values
        that are meant to be destructured immediately by the caller,
        rather than passed together as arguments to further functions.
      </p>
      <p>
        You should not return a condition object
        as one of a set of multiple values.
        Instead, you should signal the condition to denote an unusual outcome.
      </p>
      <p>
        You should signal a condition to denote an unusual outcome,
        rather than relying on a special return type.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Lists vs. Pairs">
    <SUMMARY>
      Use the appropriate functions when manipulating lists.
    </SUMMARY>
    <BODY>
      <p>
        Use <code>FIRST</code> to access the first element of a list,
        <code>SECOND</code> to access the second element, etc.
        Use <code>REST</code> to access the tail of a list.
        Use <code>ENDP</code> to test for the end of the list.
      </p>
      <p>
        Use <code>CAR</code> and <code>CDR</code>
        when the cons cell is not being used to implement a proper list
        and is instead being treated as a pair of more general objects.
        Use <code>NULL</code> to test for <code>NIL</code> in this context.
      </p>
      <p>
        The latter case should be rare outside of alists,
        since you should be using structures and classes where they apply,
        and data structure libraries when you want trees.
      </p>
      <p>
        Exceptionally, you may use <code>CDADR</code> and other variants
        on lists when manually destructuring them,
        instead of using a combination of several list accessor functions.
        In this context, using <code>CAR</code> and <code>CDR</code>
        instead of <code>FIRST</code> and <code>REST</code> also makes sense.
        However, keep in mind that it might be more appropriate in such cases
        to use higher-level constructs such as
        <code>DESTRUCTURING-BIND</code> or <code>FARE-MATCHER:MATCH</code>.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Lists vs. Arrays">
    <SUMMARY>
      You should use arrays rather than lists where random access matters.
    </SUMMARY>
    <BODY>
      <p>
        <code>ELT</code> has <i>O(n)</i> behavior when used on lists.
        If you are to use random element access on an object,
        use arrays and <code>AREF</code> instead.
      </p>
      <p>
        The exception is for code outside the critical path
        where the list is known to be small anyway.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Lists vs. Sets">
    <SUMMARY>
      You should only use lists as sets for very small lists.
    </SUMMARY>
    <BODY>
      <p>
        Using lists as representations of sets is a bad idea
        unless you know the lists will be small,
        for accessors are <i>O(n)</i> instead of <i>O(log n)</i>.
        For arbitrary big sets, use balanced binary trees,
        for instance using <code>lisp-interface-library</code>.
      </p>
      <p>
        If you still use lists as sets,
        you should not <code>UNION</code> lists just to search them.
      </p>
      <BAD_CODE_SNIPPET>
        (member foo (union list-1 list-2)) ; Bad
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        (or (member foo list-1) (member foo list-2)) ; Better
      </CODE_SNIPPET>
      <p>
        Indeed, <code>UNION</code> not only conses unnecessarily,
        but it can be <i>O(n^2)</i> on some implementations,
        and is rather slow even when it's <i>O(n)</i>.
      </p>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Proper Forms">
  <p>
    You must follow the proper usage regarding
    well-known functions, macros and special forms.
  </p>
  <STYLEPOINT title="Defining Constants">
    <SUMMARY>
      You must use proper defining forms for constant values.
    </SUMMARY>
    <BODY>
      <p>
        The Lisp system we primarily use, SBCL, is very picky and
        signals a condition whenever a constant is redefined to a value not
        <code>EQL</code> to its previous setting.
        You must not use <code>DEFCONSTANT</code>
        when defining variables that are not
        numbers, characters, or symbols (including booleans and keywords).
        Instead, consistently use whichever alternative
        is recommended for your project.
      </p>
      <BAD_CODE_SNIPPET>
        ;; Bad
        (defconstant +google-url+ "http://www.google.com/")
        (defconstant +valid-colors+ '(red green blue))
      </BAD_CODE_SNIPPET>
      
      
      <p>
        Open-Source libraries may use
        <code>ALEXANDRIA:DEFINE-CONSTANT</code>
        for constants other than numbers, characters and symbols
        (including booleans and keywords).
        You may use the <code>:TEST</code> keyword argument
        to specify an equality predicate.
      </p>
      <CODE_SNIPPET>
        ;; Better, for Open-Source code:
        (define-constant +google-url+ "http://www.google.com/" :test #'string=)
        (define-constant +valid-colors+ '(red green blue))
      </CODE_SNIPPET>
      <p>
        Note that with optimizing implementations, such as SBCL or CMUCL,
        defining constants this way precludes any later redefinition
        short of <code>UNINTERN</code>ing the symbol
        and recompiling all its clients.
        This may make it "interesting" to debug things at the REPL
        or to deploy live code upgrades.
        If there is a chance that your "constants" are not going to be constant
        over the lifetime of your server processes
        after taking into consideration scheduled and unscheduled code patches,
        you should consider using
        <code>DEFPARAMETER</code> or <code>DEFVAR</code> instead,
        or possibly a variant of <code>DEFINE-CONSTANT</code>
        that builds upon some future library implementing global lexicals
        rather than <code>DEFCONSTANT</code>.
        You may keep the <code>+plus+</code> convention in these cases
        to document the intent of the parameter as a constant.
      </p>
      <p>
        Also note that <code>LOAD-TIME-VALUE</code> may help you
        avoid the need for defined constants.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Defining Functions">
    <SUMMARY>
      You should make proper use of
      <code>&amp;OPTIONAL</code>,
      <code>&amp;KEY</code>,
      and
      <code>&amp;AUX</code> arguments.
    </SUMMARY>
    <BODY>
      <p>
        You should avoid using <code>&amp;ALLOW-OTHER-KEYS</code>,
        since it blurs the contract of a function.
        Almost any real function (generic or not) allows a certain
        fixed set of keywords, as far as its caller is concerned,
        and those are part of its contract.
        If you are implementing a method of a generic function,
        and it does not need to know
        the values of some of the keyword arguments,
        you should explicitly <code>(DECLARE (IGNORE ...))</code>
        all the arguments that you are not using.
        You must not use <code>&amp;ALLOW-OTHER-KEYS</code>
        unless you explicitly want to disable checking of allowed keys
        for all methods when invoking the generic function on arguments
        that match this particular method.
        Note that the contract of a generic function belongs in
        the <code>DEFGENERIC</code>, not in the <code>DEFMETHOD</code>
        which is basically an "implementation detail" of the generic function
        as far as the caller of the generic is concerned.
      </p>
      <p>
        A case where <code>&amp;ALLOW-OTHER-KEYS</code> is appropriate
        is when you write a wrapper function to other some other functions
        that may vary (within the computation or during development),
        and pass around a plist as a <code>&amp;REST</code> argument.
      </p>
      <p>
        You should avoid using <code>&amp;AUX</code> arguments,
        except in very short helper functions
        where they allow you to eschew a <code>LET</code>.
      </p>
      <p>
        You should avoid having both <code>&amp;OPTIONAL</code>
        and <code>&amp;KEY</code> arguments,
        unless it never makes sense to specify keyword arguments
        when the optional arguments are not all specified.
        You must not have non-<code>NIL</code> defaults
        to your <code>&amp;OPTIONAL</code> arguments
        when your function has both <code>&amp;OPTIONAL</code>
        and <code>&amp;KEY</code> arguments.
      </p>
      <p>
        For maximum portability of a library, it is good form
        that <code>DEFMETHOD</code> definitions should
        <code>(DECLARE (IGNORABLE ...))</code>
        all the required arguments that they are not using.
        Indeed, some implementations will issue a warning
        if you <code>(DECLARE (IGNORE ...))</code> those arguments,
        whereas other implementations will issue a warning
        if you fail to <code>(DECLARE (IGNORE ...))</code> them.
        <code>(DECLARE (IGNORABLE ...))</code> works on all implementations.
      </p>
      <p>
        You should avoid excessive nesting of binding forms inside a function.
        If your function ends up with massive nesting,
        you should probably break it up into several functions or macros.
        If it is really a single conceptual unit,
        consider using a macro such as <code>FARE-UTILS:NEST</code>
        to at least reduce the amount of indentation required.
        It is bad form to use <code>NEST</code> in typical short functions
        with 4 or fewer levels of nesting,
        but also bad form not to use it in the exceptional long functions
        with 10 or more levels of nesting.
        Use your judgment and consult your reviewers.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Conditional Expressions">
    <SUMMARY>
      Use the appropriate conditional form.
    </SUMMARY>
    <BODY>
      <p>
        Use <code>WHEN</code> and <code>UNLESS</code>
        when there is only one alternative.
        Use <code>IF</code> when there are two alternatives
        and <code>COND</code> when there are several.
      </p>
      <p>
        However, don't use <code>PROGN</code> for an <code>IF</code> clause
        &#8212; use <code>COND</code>, <code>WHEN</code>, or <code>UNLESS</code>.
      </p>
      <p>
        Note that in Common Lisp,
        <code>WHEN</code> and <code>UNLESS</code> return <code>NIL</code>
        when the condition is not met.
        You may take advantage of it.
        Nevertheless, you may use an <code>IF</code>
        to explicitly return <code>NIL</code>
        if you have a specific reason to insist on the return value.
        You may similarly include a fall-through clause <code>(t nil)</code>
        as the last in your <cond>COND</cond>,
        or <code>(otherwise nil)</code> as the last in your <cond>CASE</cond>,
        to insist on the fact that the value returned by the conditional matters
        and that such a case is going to be used.
        You should omit the fall-through clause
        when the conditional is used for side-effects.
      </p>
      <p>
        You should prefer <code>AND</code> and <code>OR</code>
        when it leads to more concise code than using
        <code>IF</code>, <code>COND</code>,
        <code>WHEN</code> or <code>UNLESS</code>,
        and there are no side-effects involved.
        You may also use an <code>ERROR</code>
        as a side-effect in the final clause of an <code>OR</code>.
      </p>
      <p>
        You should only use <code>CASE</code> and <code>ECASE</code>
        to compare numbers, characters or symbols
        (including booleans and keywords).
        Indeed, <code>CASE</code> uses <code>EQL</code> for comparisons,
        so strings, pathnames and structures may not compare the way you expect,
        and <code>1</code> will differ from <code>1.0</code>.
      </p>
      <p>
        You should use <code>ECASE</code> and <code>ETYPECASE</code>
        in preference to <code>CASE</code> and <code>TYPECASE</code>.
        It is better to catch erroneous values early.
      </p>
      <p>
        You should not use <code>CCASE</code> or <code>CTYPECASE</code> at all.
        At least, you should not use them in server processes,
        unless you have quite robust error handling infrastructure
        and make sure not to leak sensitive data this way.
        These are meant for interactive use,
        and can cause interesting damage
        if they cause data or control to leak to attackers.
      </p>
      <p>
        You must not use gratuitous single quotes in <code>CASE</code> forms.
        This is a common error:
      </p>
      <BAD_CODE_SNIPPET>
        (case x ; Bad: silently returns NIL on mismatch
          ('bar :bar) ; Bad: catches QUOTE
          ('baz :baz)) ; Bad: also would catch QUOTE
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        (ecase x ; Better: will error on mismatch
          ((bar) :bar) ; Better: won't match QUOTE
          ((baz) :baz)) ; Better: same reason
      </CODE_SNIPPET>
      <p>
        <code>'BAR</code> there is <code>(QUOTE BAR)</code>,
        meaning this leg of the case will be executed
        if <code>X</code> is <code>QUOTE</code>...
        and ditto for the second leg
        (though <code>QUOTE</code> will be caught by the first clause).
        This is unlikely to be what you really want.
      </p>
      <p>
        In <code>CASE</code> forms,
        you must use <code>otherwise</code> instead of <code>t</code>
        when you mean "execute this clause if the others fail".
        You must use <code>((t) ...)</code>
        when you mean "match the symbol T" rather than "match anything".
        You must also use <code>((nil) ...)</code>
        when you mean "match the symbol NIL" rather than "match nothing".
      </p>
      <p>
        Therefore, if you want to map booleans <code>NIL</code> and <code>T</code>
        to respective symbols <code>:BAR</code> and <code>:QUUX</code>,
        you should avoid the former way and do it the latter way:
      </p>
      <BAD_CODE_SNIPPET>
        (ecase x ; Bad: has no actual error case!
          (nil :bar)) ; Bad: matches nothing
          (t :quux)) ; Bad: matches anything
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        (ecase x ; Better: will actually catch non-booleans
          ((nil) :bar)) ; Better: matches NIL
          ((t) :quux)) ; Better: matches T
      </CODE_SNIPPET>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Identity, Equality and Comparisons">
    <SUMMARY>
      You should the appropriate predicates when comparing objects.
    </SUMMARY>
    <BODY>
      <p>
        Lisp provides four general equality predicates:
        <code>EQ</code>, <code>EQL</code>, <code>EQUAL</code>,
        and <code>EQUALP</code>,
        which subtly vary in semantics.
        Additionally, Lisp provides the type-specific predicates
        <code>=</code>, <code>CHAR=</code>, <code>CHAR-EQUAL</code>,
        <code>STRING=</code>, and <code>STRING-EQUAL</code>.
        Know the distinction!
      </p>
      <p>
        You should use <code>EQL</code> to compare objects and symbols
        for <em>identity</em>.
      </p>
      <p>
        You must not use <code>EQ</code> to compare numbers or characters.
        Two numbers or characters that are <code>EQL</code>
        are not required by Common Lisp to be <code>EQ</code>.
      </p>
      <p>
        When choosing between <code>EQ</code> and <code>EQL</code>,
        you should use <code>EQL</code> unless you are writing
        performance-critical low-level code.
        <code>EQL</code> reduces the opportunity
        for a class of embarrassing errors
        (i.e. if numbers or characters are ever compared).
        There may a tiny performance cost relative to <code>EQ</code>,
        although under SBCL, it often compiles away entirely.
        <code>EQ</code> is equivalent to <code>EQL</code> and type declarations,
        and use of it for optimization should be treated just like
        any such <a href="#Unsafe_Operations">unsafe operations</a>.
      </p>
      <p>
        You should use <code>CHAR=</code>
        for case-dependent character comparisons,
        and <code>CHAR-EQUAL</code> for case-ignoring character comparisons.
      </p>
      <p>
        You should use <code>STRING=</code>
        for case-dependent string comparisons,
        and <code>STRING-EQUAL</code> for case-ignoring string comparisons.
      </p>
      <p>
        A common mistake when using <code>SEARCH</code> on strings
        is to provide <code>STRING=</code> or <code>STRING-EQUAL</code>
        as the <code>:TEST</code> function.
        The <code>:TEST</code> function
        is given two sequence elements to compare.
        If the sequences are strings,
        the <code>:TEST</code> function is called on two characters,
        so the correct tests are <code>CHAR=</code> or <code>CHAR-EQUAL</code>.
        If you use <code>STRING=</code> or <code>STRING-EQUAL</code>,
        the result is what you expect,
        but in some Lisp implementations it's much slower.
        CCL (at least as of 8/2008)
        creates a one-character string upon each comparison, for example,
        which is very expensive.
      </p>
      <p>
        Also, you should use <code>:START</code> and <code>:END</code> arguments
        to <code>STRING=</code> or <code>STRING-EQUAL</code>
        instead of using <code>SUBSEQ</code>;
        e.g. <code>(string-equal (subseq s1 2 6) s2)</code> should instead be
        <code>(string-equal s1 s2 :start1 2 :end1 6)</code>
        This is preferable because it does not cons.
      </p>
      <p>
        You should use <code>ZEROP</code>,
        <code>PLUSP</code>, or <code>MINUSP</code>,
        instead of comparing a value to <code>0</code> or <code>0.0</code>.
      </p>
      <p>
        You must not use exact comparison on floating point numbers,
        since the vague nature of floating point arithmetic
        can produce little "errors" in numeric value.
        You should compare absolute values to a threshhold.
      </p>
      <p>
        You must use <code>=</code> to compare numbers,
        unless you really mean for <code>0</code>,
        <code>0.0</code> and <code>-0.0</code> to compare unequal,
        in which case you should use <code>EQL</code>.
        Then again, you must not usually use exact comparison
        on floating point numbers.
      </p>
      <p>
        Monetary amounts should be using decimal (rational) numbers
        to avoid the complexities and rounding errors
        of floating-point arithmetic.
        Libraries such as
        <a href="http://wukix.com/lisp-decimals">wu-decimal</a>
        may help you;
        once again, if this library is not satisfactory, see above about
        <a href="#Using_Libraries">Using Libraries</a> and
        <a href="#Open-Sourcing_Code">Open-Sourcing Code</a>.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Iteration">
    <SUMMARY>
      Use the appropriate form for iteration.
    </SUMMARY>
    <BODY>
      <p>
        You should simpler forms such as
        <code>DOLIST</code> or <code>DOTIMES</code>
        instead of <code>LOOP</code>
        in simple cases when you're not going to use any
        of the <code>LOOP</code> facilities such as
        bindings, collection or block return.
      </p>
      <p>
        Use the <code>WITH</code> clause of <code>LOOP</code>
        when it will avoid a level of nesting with <code>LET</code>.
        You may use <code>LET</code> if it makes it clearer
        to return one of bound variables after the <code>LOOP</code>,
        rather than use a clumsy <code>FINALLY (RETURN ...)</code> form.
      </p>
      <p>
        In the body of a <code>DOTIMES</code>,
        do not set the iteration variable.
        (CCL will issue a compiler warning if you do.)
      </p>
      <p>
        Most systems use unadorned symbols in the current package
        as <code>LOOP</code> keywords.
        Other systems use actual <code>:keywords</code>
        from the <code>KEYWORD</code> package
        as <code>LOOP</code> keywords.
        You must be consistent with the convention used in your system.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="I/O">
    <SUMMARY>
      Use the appropriate I/O functions.
    </SUMMARY>
    <BODY>
      <p>
        When writing a server,
        code must not send output to the standard streams such as
        <code>*STANDARD-OUTPUT*</code> or <code>*ERROR-OUTPUT*</code>.
        Instead, code must use the proper logging framework
        to output messages for debugging.
        We are running as a server, so there is no console!
      </p>
      <p>
        Code must not use <code>PRINT-OBJECT</code>
        to communicate with a user &#8212;
        <code>PRINT-OBJECT</code> is for debugging purposes only.
        Modifying any <code>PRINT-OBJECT</code> method
        must not break any public interfaces.
      </p>
      <p>
        You should not use a sequence of <code>WRITE-XXX</code>
        where a single <code>FORMAT</code> string could be used.
        Using format allows you
        to parameterize the format control string in the future
        if the need arises.
      </p>
      <p>
        You should use <code>WRITE-CHAR</code> to emit a character
        rather than <code>WRITE-STRING</code>
        to emit a single-character string.
      </p>
      <p>
        You should not use <code>(format nil "~A" value)</code>;
        you should use <code>PRINC-TO-STRING</code> instead.
      </p>
      <p>
        You should use <code>~&lt;Newline&gt;</code>
        or <code>~@&lt;Newline&gt;</code> in format strings
        to keep them from wrapping in 100-column editor windows,
        or to indent sections or clauses to make them more readable.
      </p>
      <p>
        You should not use <code>STRING-UPCASE</code>
        or <code>STRING-DOWNCASE</code>
        on format control parameters;
        instead, it should use <code>"~:@(~A~)"</code> or <code>"~(~A~)"</code>.
      </p>
      <p>
        Be careful when using the <code>FORMAT</code> conditional directive.
        The parameters are easy to forget.
      </p>
      <dl>
        <dt>No parameters, e.g. <code>"~[Siamese~;Manx~;Persian~] Cat"</code></dt>
        <dd>
          Take one format argument, which should be an integer.
          Use it to choose a clause. Clause numbers are zero-based.
          If the number is out of range, just print nothing.
          You can provide a default value
          by putting a <code>":"</code> in front of the last <code>";"</code>.
          E.g. in <code>"~[Siamese~;Manx~;Persian~:;Alley~] Cat"</code>,
          an out-of-range arg prints <code>"Alley"</code>.
        </dd>
        <dt><code>:</code> parameter, e.g. <code>"~:[Siamese~;Manx~]"</code></dt>
        <dd>
          Take one format argument.  If it's <code>NIL</code>,
          use the first clause, otherwise use the second clause.
        </dd>
        <dt><code>@</code> parameter, e.g. <code>"~@[Siamese ~a~]"</code></dt>
        <dd>
          If the next format argument is true,
          use the choice, but do NOT take the argument.
          If it's false, take one format argument and print nothing.
          (Normally the clause uses the format argument.)
        </dd>
        <dt><code>#</code> parameter, e.g. <code>"~#[ none~; ~s~; ~s and ~s~]"</code></dt>
        <dd>
          Use the number of arguments to format
          as the number to choose a clause.
          The same as no parameters in all other ways.
          Here's the full hairy example:
          <code>"Items:~#[ none~; ~S~; ~S and ~S~:;~@{~#[~; and~] ~S~^ ,~}~]."</code>
        </dd>
      </dl>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Optimization">
  <STYLEPOINT title="Avoid Allocation">
    <SUMMARY>
      You should avoid unnecessary allocation of memory.
    </SUMMARY>
    <BODY>
      <p>
        In a language with automatic storage management (such as Lisp or Java),
        the colloquial phrase "memory leak" refers to situation
        where storage that is not actually needed
        nevertheless does not get deallocated,
        because it is still reachable.
      </p>
      <p>
        You should be careful that when you create objects,
        you don't leave them reachable after they are no longer needed!
      </p>
      <p>
        Here's a particular trap-for-the-unwary in Common Lisp.
        If you make an array with a fill pointer, and put objects in it,
        and then set the fill pointer back to zero,
        those objects are still reachable as far as Lisp goes
        (the Common Lisp spec says that it's still OK
        to refer to the array entries past the end of the fill pointer).
      </p>
      <p>
        Don't cons (i.e., allocate) unnecessarily.
        Garbage collection is not magic.
        Excessive allocation is usually a performance problem.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Unsafe Operations">
    <SUMMARY>
      You must only use faster unsafe operations
      when there is a clear performance need
      and you can document why it's correct.
    </SUMMARY>
    <BODY>
      <p>
        Common Lisp implementations often provide backdoors
        to compute some operations faster in an unsafe way.
        For instance, some libraries provide arithmetic operations
        that are designed to be used with fixnums only,
        and yield the correct result faster if provided proper arguments.
        The downside is that the result of such operations
        is incorrect in case of overflow, and can
        have undefined behavior when called with anything but fixnums.
      </p>
      
      <p>
        More generally, unsafe operations
        will yield the correct result faster
        than would the equivalent safe operation
        if the arguments to satisfy some invariant such as
        being of the correct type and small enough;
        however if the arguments fail to satisfy the required invariants,
        then the operation may have undefined behavior,
        such as crashing the software, or,
        which is sometimes worse, silently giving wrong answers.
        Depending on whether the software is piloting an aircraft
        or other life-critical device,
        or whether it is accounting for large amounts money,
        such undefined behavior can kill or bankrupt people.
        Yet proper speed can sometimes make the difference between
        software that's unusably slow and software that does its job;
        between software that is a net loss
        and software that can yield a profit.
      </p>
      <p>
        You must not define or use unsafe operations without both
        profiling results indicating the need for this optimization,
        and careful documentation explaining why it is safe to use them.
        Unsafe operations should be restricted to internal functions;
        you should carefully documented how unsafe it is
        to use these functions with the wrong arguments.
        You should only use unsafe operations
        inside functions internal to a package and
        you should document the use of the declarations,
        since calling the functions with arguments of the wrong type
        can lead to undefined behavior.
        Use <code>check-type</code> in functions exported from a package
        to sanitize input arguments,
        so that internal functions are never passed illegal values.
      </p>
      <p>
        On some compilers,
        new unsafe operations
        can usually be defined by combining
        type declarations with an <code>OPTIMIZE</code> declaration
        that has sufficiently high <code>SPEED</code> and low <code>SAFETY</code>.
        In addition to providing more speed for production code,
        such declarations may more helpful
        than <code>check-type</code> assertions
        for finding bugs at compile-time,
        on compilers that have type inference.
        These compilers may interpret those declarations as assertions
        if you switch to safer and slower optimize settings;
        this is good to locate a dynamic error in your code during development,
        but is not to be used for production code since
        it defeats the purpose of declarations as a performance trick.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="DYNAMIC-EXTENT">
    <SUMMARY>
      You should only use <code>DYNAMIC-EXTENT</code>
      where it matters for performance,
      and you can document why it is correct.
    </SUMMARY>
    <BODY>
      <p>
        <code>DYNAMIC-EXTENT</code> declaration are
        a particular case of
        <a href="#Unsafe_Operations">unsafe operations</a>.
      </p>
      <p>
        The purpose of the <code>DYNAMIC-EXTENT</code> declaration
        is to improve performance by reducing garbage collection
        in cases where it appears to be obvious that an object's lifetime
        is within the "dynamic extent" of a function.
        That means the object is created at some point
        after the function is called, and
        the object is always inaccessible after the function exits by any means.
      </p>
      <p>
        By declaring a variable or a local function <code>DYNAMIC-EXTENT</code>,
        the programmer <em>asserts</em> to Lisp
        that any object that is ever a value of that variable
        or the closure that is the definition of the function
        has a lifetime within the dynamic extent of the (innermost) function
        that declares the variable.
      </p>
      <p>
        The Lisp implementation is then free to use that information
        to make the program faster.
        Typically, Lisp implementations can take advantage of this knowledge
        to stack-allocate:
      </p>
      <ul>
        <li>
          The lists created to store <code>&amp;REST</code> parameters.
        </li>
        <li>
          Lists and vector allocated within a function.
        </li>
        <li>
          Closures.
        </li>
      </ul>
      <p>
        If the assertion is wrong, i.e. if the programmer's claim is not true,
        the results can be <em>catastrophic</em>:
        Lisp can terminate any time after the function returns,
        or it hang forever, or &#8212; worst of all &#8212;
        produce incorrect results without any runtime error!
      </p>
      <p>
        Even if the assertion is correct,
        future changes to the function might introduce
        a violation of the assertion.
        This increases the danger.
      </p>
      <p>
        In most cases, such objects are ephemeral.
        Modern Lisp implementations use generational garbage collectors,
        which are quite efficient under these circumstances.
      </p>
      <p>
        Therefore, <code>DYNAMIC-EXTENT</code> declarations
        should be used sparingly. You must only use them if:
      </p>
      <ol>
        <li>
          There is some good reason to think that the overall effect
          on performance is noticeable, and
        </li>
        <li>
          It is absolutely clear that the assertion is true.
        </li>
        <li>
          It is quite unlikely that the code will be changed
          in ways that cause the declaration to become false.
        </li>
      </ol>
      <p>
        Point (1) is a special case of
        the principle of avoiding premature optimization.
        An optimization like this only matters if such objects
        are allocated at a very high rate, e.g. "inside an inner loop".
      </p>
      <p>
        Note that is relatively easy to ascertain that
        a function will not escape the dynamic extent of the current call frame
        by analyzing where the function is called and
        what other functions it is passed to;
        therefore, you should somewhat wary of declaring a function
        <code>DYNAMIC-EXTENT</code>, but not a high-stress declaration.
        On the other hand, it is much harder to ascertain that
        none of the objects ever bound or assigned to that variable
        will escape the dynamic extent of the current call frame,
        nor will in any future modification of a function.
        Therefore, you should be extremely wary
        of declaring a variable <code>DYNAMIC-EXTENT</code>.
      </p>
      <p>
        It's sometimes hard to know what the rate will be.
        When writing a function or macro
        that's part of a library of reusable code,
        there's no a priori way to know how often the code will run.
        Ideally, tools would be available to discover
        the availability and suitability of using such an optimization
        based on running simulations and test cases, but
        in practice this isn't as easy as it ought to be.
        It's a tradeoff.
        If you're very, very sure that the assertion is true
        (that the object is only used within the dynamic scope),
        and it's not obvious how much time will be saved
        and it's not easy to measure,
        then it may be better to put in the declaration than to leave it out.
        (Ideally it would be easier to make such measurements
        than it actually is.)
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="REDUCE vs APPLY">
    <SUMMARY>
      You should use <code>REDUCE</code>
      instead of <code>APPLY</code> where appropriate.
    </SUMMARY>
    <BODY>
      <p>
        You should use <code>REDUCE</code>
        instead of <code>APPLY</code> and a consed-up list,
        where the semantics of the first operator argument
        otherwise guarantees the same semantics.
        Of course, you must use <code>APPLY</code>
        if it does what you want and <code>REDUCE</code> doesn't.
      </p>
      <p>
        For instance, <code>(apply #'+ (mapcar #'acc frobs)</code>
        should instead be <code>(reduce #'+ frobs :key #'acc)</code>
      </p>
      <p>
        This is preferable because it does not do extra consing,
        and does not risk going beyond <code>CALL-ARGUMENTS-LIMIT</code>
        on implementations where that limit is small,
        which could blow away the stack on long lists
        (we want to avoid gratuitous non-portability in our code).
      </p>
      <p>
        However, you must be careful not to use <code>REDUCE</code>
        in ways that needlessly increase
        the complexity class of the computation.
        For instance, <code>(REDUCE 'STRCAT ...)</code> is <i>O(n^2)</i>
        when an appropriate implementation is only <i>O(n)</i>.
        Moreover, <code>(REDUCE 'APPEND ...)</code>
        is also <i>O(n^2)</i> unless you specify <code>:FROM-END T</code>.
        In such cases, you must use proper abstractions
        that cover those cases instead of calling <code>REDUCE</code>,
        first defining them in a suitable library if need be.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Avoid NCONC">
    <SUMMARY>
      You should not use <code>NCONC</code>;
      you should use <code>APPEND</code> instead,
      or better data structures.
    </SUMMARY>
    <BODY>
      <p>
        You should almost never use <code>NCONC</code>.
        You should use <code>APPEND</code>
        when you don't depend on any side-effect.
        You should use <code>ALEXANDRIA:APPENDF</code>
        when you need to update a variable.
        You should probably not depend on games
        being played with the <code>CDR</code>
        of the current CONS cell
        (which some might argue is suggested but not guaranteed by the specification);
        if you do, you must include a prominent
        comment explaining the use of <code>NCONC</code>;
        and you should probably reconsider your data representation strategy.
      </p>
      <p>
        By extension, you should avoid <code>MAPCAN</code>
        or the <code>NCONC</code> feature of <code>LOOP</code>.
        You should instead respectively use
        <code>ALEXANDRIA:MAPPEND</code>
        and the <code>APPEND</code> feature of <code>LOOP</code>.
      </p>
      <p>
        <code>NCONC</code> is very seldom a good idea,
        since its time complexity class is no better than <code>APPEND</code>,
        its space complexity class also is no better than <code>APPEND</code>
        in the common case where no one else is sharing the side-effected list,
        and its bug complexity class is way higher than <code>APPEND</code>.
      </p>
      <p>
        If the small performance hit due
        to <code>APPEND</code> vs. <code>NCONC</code>
        is a limiting factor in your program,
        you have a big problem and are probably using the wrong data structure:
        you should be using sequences with constant-time append
        (see Okasaki's book, and add them to lisp-interface-library),
        or more simply you should be accumulating data in a tree
        that will get flattened once in linear time
        after the accumulation phase is complete (see how ASDF does it).
      </p>
      <p>
        You may only use <code>NCONC</code>, <code>MAPCAN</code>
        or the <code>NCONC</code> feature of <code>LOOP</code>
        in low-level functions where performance matters,
        where the use of lists as a data structure has been vetted
        because these lists are known to be short,
        and when the function or expression the result of which are accumulated
        explicitly promises in its contract that it only returns fresh lists
        (in particular, it can't be a constant quote or backquote expression).
        Even then, the use of such primitives must be rare,
        and accompanied by justifying documentation.
      </p>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<CATEGORY title="Pitfalls">
  <STYLEPOINT title="#'FUN vs. 'FUN">
    <SUMMARY>
      You should usually refer to a function as <code>#'FUN</code> rather than <code>'FUN</code>.
    </SUMMARY>
    <BODY>
      <p>
        The former, which reads as <code>(FUNCTION FUN)</code>,
        refers to the function object, and is lexically scoped.
        The latter, which reads as <code>(QUOTE FUN)</code>,
        refers to the symbol, which when called
        uses the global <code>FDEFINITION</code> of the symbol.
      </p>
      <p>
        When using functions that take a functional argument
        (e.g., <code>MAPCAR</code>, <code>APPLY</code>,
        <code>:TEST</code> and <code>:KEY</code> arguments),
        you should use the <code>#'</code> to refer to the function,
        not just single quote.
      </p>
      <p>
        An exception is when you explicitly want dynamic linking,
        because you anticipate that
        the global function binding will be updated.
      </p>
      <p>
        Another exception is when you explicitly want to access
        a global function binding,
        and avoid a possible shadowing lexical binding.
        This shouldn't happen often, as it is usually a bad idea
        to shadow a function when you will want to use the shadowed function;
        just use a different name for the lexical function.
      </p>
      <p>
        You must consistently use either <code>#'(lambda ...)</code>
        or <code>(lambda ...)</code> without <code>#'</code> everywhere.
        Unlike the case of <code>#'symbol</code> vs <code>'symbol</code>,
        it is only a syntactic difference with no semantic impact,
        except that the former works on Genera and the latter doesn't.
        
        You must use the former style if your code is intended as a library
        with maximal compatibility to all Common Lisp implementations;
        otherwise, it is optional which style you use.
        <code>#'</code> may be seen as a hint
        that you're introducing a function in expression context;
        but the <code>lambda</code> itself is usually sufficient hint,
        and concision is good.
        Choose wisely, but above all,
        consistently with yourself and other developers,
        within a same file, package, system, project, etc.
      </p>
      <p>
        Note that if you start writing a new system
        in a heavily functional style,
        you may consider using
        <a href="http://cliki.net/lambda-reader">lambda-reader</a>,
        a system that lets you use the unicode character <code>&#955;</code>
        instead of <code>LAMBDA</code>.
        But you must not start using such a syntactic extension
        in an existing system without getting permission from other developers.
      </p>
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="Pathnames">
    <SUMMARY>
      Common Lisp pathnames are tricky. Be aware of pitfalls.
    </SUMMARY>
    <BODY>
      <p>
        It is surprisingly hard to properly deal with pathnames in Common Lisp.
      </p>
      <p>
        First, be aware of the discrepancies between
        the syntax of Common Lisp pathnames,
        which depends on which implementation and operating system
        you are using,
        and the native syntax of pathnames on your operating system.
        The Lisp syntax may involves quoting of special characters
        such as <code>#\.</code> and <code>#\*</code>, etc.,
        in addition to the quoting of
        <code>#\\</code> and <code>#\"</code> within strings.
        By contrast, your operating system's other
        system programming languages
        (shell, C, scripting languages)
        may only have one layer of quoting, into strings.
      </p>
      <p>
        Second, when using <code>MERGE-PATHNAMES</code>,
        be wary of the treatment of the <code>HOST</code> component,
        which matters a lot on non-Unix platforms
        (and even on some Unix implementations).
        You probably should be using
        <code>ASDF-UTILS:MERGE-PATHNAMES*</code>
        instead of <code>MERGE-PATHNAMES</code>,
        especially if your expectations for relative pathnames
        are informed by the way they work in Unix or Windows;
        otherwise you might hit weird bugs whereby on some implementations,
        merging a relative pathnames with an absolute pathname
        results in overriding the absolute pathname's host
        and replace it with the host from the value of
        <code>*DEFAULT-PATHNAME-DEFAULTS*</code>
        at the time the relative pathname was created.
      </p>
      <p>
        Third, be aware that <code>DIRECTORY</code>
        is not portable across implementations
        in how it handles wildcards, sub-directories, symlinks, etc.
        There again, <code>ASDF-UTILS</code> provides several
        common abstractions to deal with pathnames.
      </p>
      <p>
        <code>LOGICAL-PATHNAME</code>s are not a portable abstraction,
        and should not be used in portable code.
        Many implementations have bugs in them, when they are supported at all.
        SBCL implements them very well,
        but strictly enforces the limitations on characters
        allowed by the standard.
        Other implementations allow arbitrary characters in such pathnames,
        but in doing so are not being conformant,
        and are still incompatible with each other in many ways.
        You should use other pathname abstractions,
        such as <code>ASDF:SYSTEM-RELATIVE-PATHNAME</code> or
        the underlying <code>ASDF-UTILS:COERCE-PATHNAME</code>.
      </p>
      
      <p>
        Finally, be aware that paths may change between
        the time you build the Lisp image for your application,
        and the time you run the application from its image.
        You should be careful to reset your image
        to forget irrelevant build-time paths and
        reinitialize any search path from current environment variables.
        <code>ASDF</code> for instance requires you to reset its paths
        with <code>ASDF:CLEAR-CONFIGURATION</code>.
      </p>
      
    </BODY>
  </STYLEPOINT>
  <STYLEPOINT title="SATISFIES">
    <SUMMARY>
      You must be careful when using a <code>SATISFIES</code> clause in a type specifier.
    </SUMMARY>
    <BODY>
      <p>
        Most Common Lisp implementations can't optimize
        based on a <code>SATISFIES</code> type,
        but many of them offer simple optimizations
        based on a type of the form
        <code>(AND FOO (SATISFIES BAR-P))</code>
        where the first term of the <code>AND</code> clause
        describes the structure of the object
        without any <code>SATISFIES</code>
        and the second term is the <code>SATISFIES</code>.
      </p>
      <BAD_CODE_SNIPPET>
        (deftype prime-number () (satisfies prime-number-p)) ; Bad
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        (deftype prime-number () (and integer (satisfies prime-number-p)) ; Better
      </CODE_SNIPPET>
      <p>
        However, <code>AND</code> in the <code>DEFTYPE</code> language
        isn't a left-to-right short-circuit operator
        as in the expression language;
        it is a symmetrical connector that allows for reordering subterms
        and doesn't guarantee short-circuiting.
        Therefore, in the above example,
        you cannot rely on the test for <code>INTEGER</code>ness
        to protect the function <code>PRIME-NUMBER-P</code>
        from being supplied non-integer arguments
        to test for being of instances of the type.
        Implementations may, and some <em>will</em>,
        invoke <code>SATISFIES</code>-specified function
        at compile-time to test various relevant objects.
      </p>
      <p>
        That is why any function specified in a <code>SATISFIES</code> clause
        MUST accept objects of any type as argument to the function,
        and MUST be defined within an <code>EVAL-WHEN</code>.
      </p>
      <BAD_CODE_SNIPPET>
        (defun prime-number-p (n) ; Doubly bad!
          (let ((m (abs n)))
            (if (&lt;= m *prime-number-cutoff*)
                (small-prime-number-p m)
                (big-prime-number-p m))))
      </BAD_CODE_SNIPPET>
      <CODE_SNIPPET>
        (eval-when (:compile-toplevel :load-toplevel :execute) ; Better
          (defun prime-number-p (n)
            (when (integerp n) ; Better
              (let ((m (abs n)))
                (if (&lt;= m *prime-number-cutoff*)
                    (small-prime-number-p m)
                    (big-prime-number-p m))))))
      </CODE_SNIPPET>
      <p>
        In particular, the above means that the
        <a href="http://www.lispworks.com/documentation/HyperSpec/Body/t_satisf.htm">example</a>
        used in the Common Lisp Standard is erroneous:
        <code>(and integer (satisfies evenp))</code>
        is <em>not</em> a safe, conformant type specifier to use,
        because <code>EVENP</code> will throw an error
        rather than return <code>NIL</code>
        when passed a non-integer as an argument.
      </p>
      <p>
        Finally, there is a catch when your <code>DEFTYPE</code> code expands
        to a <code>SATISFIES</code> with a dynamically generated function:
      </p>
      <ul>
        <li>
          You cannot control when implementations will or will not
          expand a <code>DEFTYPE</code>.
        </li>
        <li>
          The expansion itself cannot contain a function definition
          or any code in the expression language.
        </li>
        <li>
          You cannot control when the expansion is used,
          it may happen in a different process
          that didn't expand the definition.
        </li>
      </ul>
      <p>
        Therefore, you cannot merely create the function
        as a side-effect of expansion
        using <code>EVAL</code> at type-expansion time.
        The solution is to use
        <code>ASDF-FINALIZERS:EVAL-AT-TOPLEVEL</code> instead.
        See the very last point
        in the discussion about <a href="#EVAL">EVAL</a>.
      </p>
      <p>
        Common Lisp is hard to satisfy.
      </p>
    </BODY>
  </STYLEPOINT>
</CATEGORY>

<HR/>

<small>Credits:
   Adam Worrall, Dan Pierson, Matt Marjanovic, Matt Reklaitis,
   Paul Weiss, Scott McKay, Sundar Narasimhan,
   and several other people contributed.
   Special thanks to Steve Hain,
   and to the previous editors,
   in reverse chronological order Dan Weinreb and Jeremy Brown.
</small>

<p align="right">
Revision 1.17
</p>


<address>
Robert Brown
</address>

<address>
  <a HREF="mailto:tunes@google.com">Fran&#231;ois-Ren&#233; Rideau</a>
</address>


</GUIDE>