Unfortunately that time of year has crept up… school has started again. And not just any kind of school: I find myself in 12th grade, balancing seven classes, two with AP tests at the end, three with huge final projects. I’m also trying to apply to college at …
For the past week or two I have been working on a few things simultaneously. I keep prototype code that I’m developing but not quite ready to release yet in private GitHub repositories, and publish them when I’m ready. Lately I’ve been working on two different and …
Page 1 of 5 Next> diff --git a/docs/index_2.html b/docs/index_2.html index af16010..1e4e167 100644 --- a/docs/index_2.html +++ b/docs/index_2.html @@ -74,6 +74,8 @@
For the past week or two I have been working on a few things simultaneously. I keep prototype code that I’m developing but not quite ready to release yet in private GitHub repositories, and publish them when I’m ready. Lately I’ve been working on two different and …
Some time ago I ordered one of Parallax’s “Feedback 360” servos for a robotics project I’m working on. I wrote a little arduino library to control it, and it works well.
-As-is, however, the servo itself is too big to fit where I want it to. I’m …
<<First Page 2 of 5 diff --git a/docs/index_3.html b/docs/index_3.html index c510552..1d139c6 100644 --- a/docs/index_3.html +++ b/docs/index_3.html @@ -74,6 +74,8 @@
Some time ago I ordered one of Parallax’s “Feedback 360” servos for a robotics project I’m working on. I wrote a little arduino library to control it, and it works well.
+As-is, however, the servo itself is too big to fit where I want it to. I’m …
Aaah, that’s been a nice break from coding. Four weeks with no commits pushed to Phoo. It’s a shame I left it in a broken state. I haven’t the slightest idea why it’s broken — but then again, all the other times Phoo broke, I never knew …
<<First <Previous diff --git a/docs/index_4.html b/docs/index_4.html index 01b5fda..6c53a0d 100644 --- a/docs/index_4.html +++ b/docs/index_4.html @@ -74,6 +74,8 @@
Aaah, that’s been a nice break from coding. Four weeks with no commits pushed to Phoo. It’s a shame I left it in a broken state. I haven’t the slightest idea why it’s broken — but then again, all the other times Phoo broke, I never knew …
After several days trying to debug the horribly bug-ridden local-echo package for xterm.js, I have had enough. It’s Ioannis’ responsibility to debug his package. I am switching over to another Javascript terminal ptogram I found, jquery.terminal by Jakub T. Jankiewicz. It has built-in readline support, and well …
<<First <Previous diff --git a/docs/index_5.html b/docs/index_5.html index 15c5e9b..fddaeb4 100644 --- a/docs/index_5.html +++ b/docs/index_5.html @@ -74,6 +74,8 @@
After several days trying to debug the horribly bug-ridden local-echo package for xterm.js, I have had enough. It’s Ioannis’ responsibility to debug his package. I am switching over to another Javascript terminal ptogram I found, jquery.terminal by Jakub T. Jankiewicz. It has built-in readline support, and well …
Unfortunately that time of year has crept up… school has started again. And not just any kind of school: I find myself in 12th grade, balancing seven classes, two with AP tests at the end, three with huge final projects. I’m also trying to apply to college at the same time. Not to mention that I also have a bunch of my own personal projects that I’m trying to work on too.
+Thankfully I have had a small amount of time to spend on planning a couple of different things.
+I completed and published Parasite last week. It’s not terribly polished and the game is practically unplayable, but I am still pretty impressed with where it is so far. It’s a bit of an interesting experiment to see what you can do with a bunch of completely untrained neural networks. I encourage you to try it out and tell me what you think.
+The last thing I want to do but can’t at this moment. The game is currently playable using a keyboard and mouse, but I wanted to also be able to add gamepad support. I don’t own any gamepads, so that’s going to have to wait.
+Another idea I considered would be to add a “sandbox” level at the end, in which the player would have more freedom to do anything they wanted, with no “goal” to determine when the level is complete. I haven’t implemented any of this yet but I might. Nothing is certain at this point.
+Not much has happened with PICKLE since my last post about it. However, in my investigation of other things I came up wth a few more substantial ideas, which are outlined below.
+Currently the indentation feature is triggered by the sequence of colon plus newline plus an indented block, like this:
+## This parses as a colon block string, and a single line.
+foo bar:
+ baz bam
+
+## This causes (or will cause) an "unexpected unindent" error.
+foo bar
+ baz bamThe mandatory colon has been haunting me for a while and I finally figured out how to remove it.
+Normally, the parser “looks ahead” to check if there is a colon block and if there is it removes the colon from the previous token. However, I realized that it would still be able to detect the indent regardless: when it is parsing the colon block it strips the indent from all of the lines. Functions that take blocks of code instead of strings will trigger the colon block to be parsed, and by then the indent will be gone, so the parser only needs to care about if there is an indent, not the size of the indent (just enough to verify that it’s all valid).
+When it detects a newline, it looks to see if the next line is indented, and if so, it takes back the newline token and starts chucking the (stripped) lines into a buffer. When it detects a line with no indent, it emits the indent-block string onto the same line as the block header and then finally emits a newline.
+Conveniently, this also fixes a problem where the parser wouldn’t accept a comment between the colon and the newline when opening a colon-block, because now the colon is just another operator.
+While working on parsers for some other mini-languages I discovered an interesting problem that arises when the language includes alphabetic operators such as and and or and also allowes unquoted strings (there are no symbols). The problem arose when an unquoted string contained a keyword: if I wrote sandbox world, it would actually be interpreted as if I had typed s and box w or ld. This caused a lot of head-scratching until I figured out what was going on and quoted the strings.
PICKLE has symbols, instead of unquoted strings, but the problem could still happen with the every-character-is-its-own-symbol rule I presented previously. This would be solved immediately if I automatically concatenated symbols in the same class to each other before any rules were applied. The most basic rules I can think of would be categories of letters and underscores, digits, and punctuation characters (excluding parentheses which are handled specially by the parser).
+The next step could be to add patterns for different number formats, etc. but I want to avoid locking-in specific formats. One of the goals of PICKLE is to have no fixed formats and that would go against it. The easiest solution to this would be to combine PICKLE’s pattern matching with a feature of Tcl: each token always has a string representation. That way, when tokens match concatenation rules they could dig up the original string that the programmer wrote and parse that instead of trying to muck around with the object that the token turned into. Better yet, each token could also remember where it was originally written, to be able to give helpful error messages.
+In playing around with the Scheme programming language, I had a few ideas. Compliant Scheme systems implement everything in continuation-passing style. That is, instead of simply returning a result to the caller, each operation calls the current continuation with the result. If the operation could not be completed, it calls the failure continuation with the details of the error.
+Continuations are best illustrated, I think, using Javascript Promises. They are, in effect, a limited form of call-with-current-continuation-and-current-failure-continuation:
new Promise((resolve, reject) => {
+ // resolve is the current continuation
+ // reject is the failure continuation
+});Scheme also has a construct known as dynamic-wind that functions a lot like Python’s with blocks. dynamic-wind takes three thunks, the first is called whenever code in the second starts running, and the third is called whenever code in the second stops running. Entry and exit in Python is only possible by way of normal control flow or thrown exceptions, but in Scheme it’s possible to ender and exit the same block many times, using continuations.
The idea I had came after seeing a little snippet demonstrating the use of SISC Scheme’s with-failure-continuation, which acts almost like a Python try block, except for the fact that it is also passed the current continuation at the point where the error occurred. This is extremely powerful.
Consider a simple divide-by-zero error. The current continuation of the division expression still exists, and now the failure handler can decide whether to resume or propagate the exception. Python doesn’t let you decide what to do in this situation: the result is always an exception being thrown. However, in Scheme, the failure handler can instead resume the computation with a substitute value: in some situations it would be appropriate to return $\pm\infty$ for a division by zero; in other situations a NaN value; and in others to continue the error.
When errors are resumable, it unlocks a lot of possibilities. One useful example I can think of is quantities (numbers with units): a quantities package could install a failure handler that would intercept errors caused by trying to call a number with arguments (which normally would make no sense) and resume the call with a quantity.
+For example (10 'miles) normally would normally not work (Guile gives the message “Wrong type to apply: 10”), but if the error included details of what was being called and with what arguments, a handler could transform this call into a quantity representing a distance of 10 miles.
So I’ve decided that in PICKLE, continuations will be supplied with every error, so the computation can be resumed with any specific values.
+That’s all for now. I hope to be able to implement some of this soon!
+Related Posts
+