Skip to content

Latest commit

 

History

History
375 lines (233 loc) · 15.2 KB

CHANGELOG.md

File metadata and controls

375 lines (233 loc) · 15.2 KB
Raw

Changelog

2.1.0

Thank you https://github.com/zediogoviana, for the following improvements:

  • Add configurable SSL options
  • Fix local connection
  • Keep the :ssl keyword to manage the options also

Dependencies updated, paving the road to switching to latest Elixir/Erlang combo.

2.0.11

  • Issue #100: Timeout set in config in now used by queries
  • DBConnection, bump dependencies

2.0.10

  • Fix temporal types usage: microseconds are not fully available
  • Review to pass test on Neo4j 4 :
    • test and doctests to use new parameter syntax (using {} is deprecated in Neo4j 4)
    • toUpper instead of upper

2.0.9

2.0.8

  • Fix Response.profile not being properly filled. Closes #91

2.0.7

  • sometimes the server version is missing the patch number, and the router couldn't return the proper version. Thank you @barry-w-hill, for finding this bug and reporting it!
  • remove the basic_auth when using the &Bolt.Sips.info/0 function. Thanks @dominique-vassard, for suggestion. Closes: #89

2.0.6

  • Fix 'unused alias' compilation warnings
  • Fix Bolt.Sips.Response type: stats was a list instead of list|map
  • Add typespec for Bolt.Sips.Types: Node, Relationship and UnboundRelationship

2.0.5

2.0.4

  • feature: support connection options in queries PR #82. Many thanks @tcrossland, for this contribution! This PR adds support for passing options through to DBConnection.execute/4
  • fix some broken links, in the docs; closes #76
  • update some dependencies, including the DBConnection package.
  • squashing some compile warnings; to be continued /attn: @team ;)
  • please use Elixir 1.9 or 1.10, for test and development - where possible.

2.0.3

  • refactoring the internals for achieving a better performance, while improving the code readability and extensibility - many thanks to @kristofka and @dominique-vassard. You guys are awesome!
  • fix: Consistent bad connection state after malformed query [...] issue #78

=== 2.0.2 ===

  • The 2.0, stable release. Thank you all for your feedback and for contributing to making this driver better. w⦿‿⦿t!
  • fix: Simple Query taking too much time to process #73

2.0.0-rc.2

  • swapping the assets around, for better organizing the docs

2.0.0-rc.1

  • more documentation
  • fix the TravisCi build
  • min versions erlang 21.2 elixir 1.7

=== 2.0.0-rc ===

What's New?

bolt+routing:// is now supported

Read more what this schema is, as defined by the Neo4j team

Role-based connections

Until this version, Bolt.Sips was used for connecting to a single Neo4j server, aka: the "direct" mode. Basically you configure the driver with a url to a Neo4j server and Bolt.Sips will use that to attach itself to it, using a single configuration, remaining attached to that server until it is restarted (or reconfigured). In direct mode, bolt_sips "knows" only one server.

Starting with this version you can have as many distinct connection configurations, each of them dedicated to different Neo4j servers, as/if needed. We call these connections: "role-based connections". For example, when you'll connect to a Neo4j cluster using the new protocol, i.e. by using a configuration like this:

config :bolt_sips, Bolt,
  # default port considered to be: 7687
  url: "bolt+routing://localhost",
  basic_auth: [username: "neo4j", password: "test"],
  pool_size: 10

Bolt.Sips will automatically create three pools of size 10, with the following reserved names: :read, :write and :route. Now you can specify what type of connection you want to use, by its name (role). For example:

wconn = Bolt.Sips.conn(:write)
... = Bolt.Sips.query!(wconn, "CREATE (a:Person {name:'Bob'})")

rconn = Bolt.Sips.conn(:read)
... = Bolt.Sips.query!(rconn, "MATCH (a:Person {name: 'Bob'}) RETURN a.name AS name")

The roles above: :read, :write and :route, are reserved. Please do not name custom connections using the same names (atoms). And as you just realized, yes: now you can create as many Bolt.Sips direct "driver instances" as you want, or as many as your app/hardware supports.

Please see the documentation for much more details.

Main breaking changes introduced in version 2.x

  • the hostname config parameter is a string; used to be a charlist

  • the url config parameter must start with a valid schema i.e. bolt, bolt+routing or neo4j. Examples:

    url: "bolt://localhost"
url: "bolt+routing://neo4j:[email protected]:123456?policy=europe"

  • Bolt.Sips.Query, will return a Bolt.Sips.Response now; it used to be a simple data structure.

=== 1.5 ===

1.5.1

  • add a test alias for running the tests compatible with the most recent Neo4j server while disabling the older/legacy ones
  • cleanup some warning about unused aliases

1.5.0

  • Bolt V3 support
  • Decompose tests by bolt version
  • Important note about transaction

1.4.0

  • Encoding / Decoding types is now at the lowest possible level
  • Decompose encoders / decoders by bolt version
  • Expose only public API in docs

1.3.0

  • 1.3.0 stable release. Many thanks to Dominique VASSARD, for his awesome contributions.

1.3.0-rc2

  • Fix some typos
  • add json encoding capability

1.2.2-rc2

  • Bug fix: Nanoseconds formating was erroneous. Example: 54 nanoseconds was formated to "PT0.54S" instead of "PT0.000000054S"
  • Bug fix: Large amount of nanoseconds (>= 1_000_000) wasn't treated and lead to Neo4j errors. Now large amount of nanoseconds are converted in seconds, with the remainder in nanoseconds.

1.2.1-rc2

  • Bug fix: If a property contains a speciifc types (date, datetime, point, etc.), it wasn't decoded. see: florinpatrascu#55

1.2.0-rc2

  • support for the spatial and temporal types.

1.1.0-rc2

  • removed the boltex dependency and added all its "low-level" code to internals.

1.0.0-rc2

Breaking changes introduced in version 1.x

  • non-closure based transactions are not supported anymore. This is a change introduced in DBConnection 2.x. Bolt.Sips version tagged v0.5.10 is the last version supporting open transactions.
  • the support for ETLS was dropped. It was mostly used for development or hand-crafted deployments

This version is using the official DBConnection 2.0.0-rc2, from hex.pm

0.5.10

  • update the links referencing the Bolt protocol documentation (types, etc)

0.5.9

  • upgrade dependencies
  • trading carefully around the new db_connection, as we're chasing the code from master currently, and there more changes in the pipe to come for the both projects; db_connection, and this one, respectively.

0.5.8

  • dealing with negative integers see issue #42, for more details

0.5.7

  • elixir 1.6 and code formatting, of course :)
  • minor test updates
  • update dependencies
  • pending code for the newest db_connection (currently using db_connection from the master branch)

0.5.5

  • using the DBConnection, thanks to the work done by Dmitriy Nesteryuk.

0.4.11

  • using Elixir 1.5
  • not using the ConCache anymore. I initially intended to use its support throughout the driver, but it is not needed.
  • README updated with a short snippet from a Phoenix web app demo, showing how to start Bolt.Sips, as a worker
  • dependencies update
  • minor code cleanup, to prep the code for receiving HA and Bolt routing capabilities

v0.3.5

  • better error messages; issue #33
  • not retrying a connection when the server is not available/started
  • incorrect number of retries, performed by the driver in case of errors; was one extra

v0.3.4

  • dependencies update, minor code cleanup, listening to Credo :) and finally using a Markdown linter

v0.3.3

  • Add link to travis build; #31 by vic

v0.3.2

  • Use the project's own configuration file when executing the bolt.cypher mix task. Fixes issue #20

v0.3.1 Breaking changes

  • rollback/refactor to optionally allow external configuration options to be defined at runtime. You must start the Bolt.Sips manually, when needed, i.e. Bolt.Sips.start_link(url: "localhost"), or by changing your app's mix config file, i.e.
def application do
  [applications: [:logger, :bolt_sips],
   mod: {Bolt.Sips.Application, []}]
end

You can also specify custom configuration settings in you app's mix config file. These may overwrite your config file:

def application do
  [extra_applications: [:logger], mod:
    {Bolt.Sips.Application, [url: 'localhost', pool_size: 15]}
  ]
end
  • code cleanup

v0.2.6 (2017-04-21)

  • cleanup, and minor dependencies update

v0.2.5 (2017-03-22)

  • split multi-line Cypher statements containing semicolons only if the ; character is at the end of the line, followed by \r\n on Windows and \n on Unix like system, otherwise it may break the Cypher statement when the semicolon appears somewhere else

v0.2.4 (2017-02-26)

  • add the fuzzyurl to the list of apps, for project using Elixir < 1.4 (thank you, @dnesteryuk!)

v0.2.3 (2017-02-26)

  • improved connection handling

v0.2.2 (2017-02-24)

  • PR #18; Bring up :boltex and :retry in applications, for Elixir < 1.4 (from: @wli0503, thank you!)
  • PR #19; test for error message on invalid parameter types (from: @vic, thank you!).

v0.2.1 (2017-02-20)

  • stop retrying a request if the failure is an internal one (driver, or driver dependencies related).
  • update the Boltex driver containing two important bug fixes: one where Boltex will fail when receiving too much data (florinpatrascu/issues/16) and the other one, an improvement, make Boltex.Error.get_id/1 more resilient for new transports (details here: mschae/boltex/issues/14)
  • changed the pool strategy to :fifo, and its timeout to :infinity, and let the (:gen_server) call timeout expire according to the user's :timeout configuration parameter
  • added a test unit provided by @adri (thank you), for executing a Cypher query, with large set of parameters

v0.2.0 Breaking changes

  • Elixir 1.4 is now required.
  • Using Boltex 0.2.0
  • bugfix: invalid Cypher statements will now be properly handled when the request is retried automatically

v0.1.11

  • With a larger amount of parameters it seems like generating chunks isn't working correctly. This is a patch imported from Boltex, see: mschae/boltex#13, for more info

v0.1.10 (2017-02-11)

  • accept Map and Struct for query parameters, transparently. Thank you [@wli0503], for the PR.

v0.1.9 (2017-01-27)

Some of the users are encountering difficulties when trying to compile bolt_sips on Windows. This release is addressing their concern.

Bolt.Sips will use the optional System variable: BOLT_WITH_ETLS, for depending on the ETLS package. If that variable is not defined, then Bolt.Sips will use the standard Erlang :ssl module, for the SSL/TLS protocol; the default behavior, starting with this version.

Therefore, if you want the much faster ssl/tls support offered by ETLS, then use this: export BOLT_WITH_ETLS=true on Linux/OSX, for example. Then:

 mix deps.get
 mix test

and so on.

(Don't forget to mix deps.unlock --all, if you plan to plan to further debugging/developing w/ or w/o the BOLT_WITH_ETLS support)

Many thanks to: Ben Wilson, for advices.

v0.1.8 (2017-01-07)

  • using Elixir 1.4
  • add more details to the README, about the components required to build ETLS, the TCP/TLS layer
  • added newer Elixirs to the Travis CI configuration file
  • minor code cleanups

v0.1.7 (2017-01-02)

  • Connection code refactored for capturing the errors when the remote server is not responding on the first request, or if the driver is misconfigured i.e. wrong port number, bad hostname ...
  • updated the test configuration file with detailed info about the newly introduced option: :retry_linear_backoff, mostly as a reminder

v0.1.6 (2017-01-01)

  • we're already using configurable timeouts, when executing requests from the connection pool. But with Bolt, the initial handshake sequence (happening before sending any commands to the server) is represented by two important calls, executed in sequence: handshake and init, and they must both succeed, before sending any (Cypher) requests. You can see the details in the Bolt protocol specs. This sequence is also sensitive to latencies, such as: network latencies, busy servers, etc., and because of that we're introducing a simple support for retrying the handshake (and the subsequent requests) with a linear backoff, and try the handshake sequence (or the request) a couple of times before giving up - all these as part of the exiting pool management, of course. This retry is configurable via a new configuration parameter, the: :retry_linear_backoff, respectively. For example:
config :bolt_sips, Bolt,
  url: "bolt://Bilbo:[email protected]:24786",
  ssl: true,
  timeout: 15_000,
  retry_linear_backoff: [delay: 150, factor: 2, tries: 3]

In the example above the retry will linearly increase the delay from 150ms following a Fibonacci pattern, cap the delay at 15 seconds (the value defined by the :timeout parameter) and giving up after 3 attempts. The same retry mechanism (and configuration parameters) is also honored when we send requests to the neo4j server.

v0.1.5 (2016-12-30)

  • as requested by many users, this version is introducing the optional url configuration parameter. If present, it will be used for extracting the host name, the port and the authentication details. Please see the README, for a couple of examples. For brevity:
config :bolt_sips, Bolt,
  url: 'bolt://demo:[email protected]:24786',
  ssl: true

v0.1.4 (Merry Christmas)

  • add support for connecting to Neo4j servers on encrypted sockets. Currently only TLSv1.2 is supported, using the default BoringSSL cipher; via :etls. To connect securely to a remote Neo4j server, such as the ones provided by graphenedb.com, modify your Bolt.Sips config file like this (example):
config :bolt_sips, Bolt,
  hostname: 'bolt://hobby-blah.dbs.graphenedb.com',
  basic_auth: [username: "wow", password: "of_course_this_is_the_password"],
  port: 24786,
  pool_size: 5,
  ssl: true,
  max_overflow: 1

Observe the new flag: ssl: true

Please note this is work in progress

v0.1.2 (2016-11-06)

  • integrate the Boltex code from https://github.com/mschae/boltex, and let the Bolt.Sips wrapper to manage the connectivity, using a simple Poolboy implementation for connection pooling

v0.1.1 (2016-09-09)

  • a temporary solution for dealing with negative values while extracting a graph walk-through from a Path. Dealing with this in Boltex instead, but this fix should work for now.

v0.1.0 (2016-08-31)

First release!