New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Testing removal of 'skim' #2520

Merged
merged 2 commits into from May 29, 2018

Conversation

Projects
None yet
4 participants
@wubzz
Collaborator

wubzz commented Feb 28, 2018

Temporary, might be closed.

wubzz added some commits Feb 28, 2018

@wubzz

This comment has been minimized.

Collaborator

wubzz commented Feb 28, 2018

@tgriesser This function is 4 years old and hasn't been changed. This is a long shot in the dark, but do you happen to remember the reasoning behind it? 😄 It essentially ensures returned rows do not return any non-enumerable properties.

No tests are failing after removing it. Still not sure if that's reliable enough, tho.

Currently this is a progressive performance bottleneck as described in #2518 (comment)

@wubzz wubzz requested review from tgriesser and elhigu Feb 28, 2018

@elhigu

This comment has been minimized.

Collaborator

elhigu commented Feb 28, 2018

Maybe it converts random classes etc. returned from drivers to plain objects?

@elhigu

This comment has been minimized.

Collaborator

elhigu commented Feb 28, 2018

Anyways I'm pretty if it is wanted then it would be faster to do with plain javascript loop. @koskimas has done lots of micro optimizations to obejction, he probably remembers from top of the head if there is something really inefficient in that implementation.

@tgriesser

This comment has been minimized.

Owner

tgriesser commented Feb 28, 2018

Maybe it converts random classes etc. returned from drivers to plain objects?

Exactly it. I forget which drivers it was now though

@wubzz

This comment has been minimized.

Collaborator

wubzz commented Feb 28, 2018

That makes sense.

Pressumably the only reason it affects so many dialects is because copy-paste when adding new dialects. I imagine mssql can be one such case.

@wubzz wubzz closed this Feb 28, 2018

@elhigu

This comment has been minimized.

Collaborator

elhigu commented Feb 28, 2018

Looks like this could be used also to copy object own properties to plain object https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Object/assign

If I didn't do something badly wrong with object assign skimming is ~20x faster https://jsperf.com/knex-skim (on chrome at least)

That wouldn't be anymore noticeable difference.

@koskimas

This comment has been minimized.

Collaborator

koskimas commented Feb 28, 2018

@elhigu I was about to suggest the same. V8 team has spent some time optimizing Object.assign recently. It will probably get even faster with node 10.

@koskimas

This comment has been minimized.

Collaborator

koskimas commented Feb 28, 2018

I think you could even skip the whole thing if obj.constructor === Object. Even thought it is fast enough, that creates a copy of each object producing unnecessary load for the gc. But once again, premature optimization is the root of all evil and all that 😄

@tgriesser

This comment has been minimized.

Owner

tgriesser commented May 2, 2018

Actually, I just looked into it cause I had some perf issues myself with this. I believe it can be removed in mysql, since this fixes its original use (only enumerating fields):

mysqljs/mysql@9cf7153

Didn't look into other dialects yet.

@tgriesser tgriesser reopened this May 2, 2018

@wubzz

This comment has been minimized.

Collaborator

wubzz commented May 23, 2018

@tgriesser @elhigu I'm comfortable removing this for both MySQL (from Tim's comment) and Postgres. I can't speak for other dialects, tho. Anything we can do to speed up resultset is a win-win situation imo. If you're ok with this I can update the PR.

Edit: Heh, isn't used for Postgres, so nevermind on pg. :)

@elhigu

This comment has been minimized.

Collaborator

elhigu commented May 24, 2018

@wubzz I believe this is good to go, just need to remember write to changelog about that returned row types might not be plain js objects anymore on those dialects, which were skimmed earlier.

@tgriesser tgriesser merged commit cf51f5c into tgriesser:master May 29, 2018

1 check passed

continuous-integration/travis-ci/pr The Travis CI build passed
Details

veikovx added a commit to veikovx/knex that referenced this pull request Sep 17, 2018

elhigu added a commit that referenced this pull request Sep 26, 2018

Kill queries after timeout for PostgreSQL (#2636)
* Kill queries after timeout for PostgreSQL

* Fix cancellation connection acquiring and test.

* Fix releasing connection in case query cancellation fails

* Add support for native enums on Postgres (#2632)

Reference https://www.postgresql.org/docs/current/static/sql-createtype.html

Closes #394

Signed-off-by: Will Soto <will.soto9@gmail.com>

* Removal of 'skim' (#2520)

* Allow overwriting log functions (#2625)

* Example build of custom log functions
* Handle logger object better for transactions
* Adjust test to ignore sqlite warning message

* Fixed onIn with empty values array (#2513)

* Drop support for strong-oracle (#2487)

* Remove babel-plugin-lodash (#2634)

While in theory, this may reduce the bundle size,
in practice it adds a ton of overhead during startup
due to the number of additional requires. Bundle
size also shouldn't matter for server side modules.

* Add json support to the query builder for mysql (#2635)

* Add json support to the query builder for mysql

refs #1036

Based on #1902

* Clarify supported version

* fix wrapIdentifier not being called in postgres alter column (#2612)

* fix wrapIdentifier not being called in postgres alter column

* add test for wrapIdentifier call in postgres alter column

* add comment regarding issue

* add issue & PR #'s in comment

* Remove readable-stream and safe-buffer (#2640)

* Add json support to the query builder for mysql

refs #1036

Based on #1902

* Clarify supported version

* Use native Buffer and Stream implementations

* fixes 2630 (#2642)

* Timeout errors shouldn't silently ignore the passed errors, but rather reject with original error. Fixes #2582 (#2626)

* chore: add Node.js 10 (#2594)

* chore: add Node.js 10

* chore: trigger new build

* chore: update lockfile

* chore: trigger new build

* fix: use npm i instead of npm ci

* Fix mssql driver crashing (#2637)

* Run SQL Server tests locally running SQL server in docker

* WIP mssql test stuff

* Patched MSSQL driver to not crash knex anymore

* Removed semicolon from rollback stmt for oracle (#2564)

* Remove WebSQL dialect (#2647)

* Add json support to the query builder for mysql

refs #1036

Based on #1902

* Clarify supported version

* Use native Buffer and Stream implementations

* Remove WebSQL dialect

* add homepage field to package.json (#2650)

* Make the stream catch errors in the query (#2638)

* Make the stream catch errors in the query

* Fix another case in which stream doesnt emits error

* Linter stuff

* Remove setTimeout in tests

* Make a test not to check the MySQL error code

* Fix stream error catching for MariaDB and PostgreSQL

* Fix stream error catching in Oracle

* Throw the error after emitting it to the stream

* Throw the error without instantiating a new Error

* Various fixes to mssql dialect (#2653)

* Fixed float type of mssql to be float

* Many tests where postgres test was not actually ran at all

* Migrations to be mssql compatible

Mssql driver doesn't handle if multiple queries are sent to same transaction concurrently.

* Prevented mssql failing when invalid schema builder was executed by accident

Instead of trying to generate sql from broken schema calls, just make exception to leak before query compiling is started

* Fixed mssql trx rollback to always throw an error

Also modified some connection test query to be mssql compatible

* Fixed various bugs from MSSQL driver to make tests run

* Fixed mssql unique index to be compatible with other dialect implementations

* Enable running mssql tests on CI

* Test for #2588

* Updated tests to not be dependend on tables left from previous test rans

* Trying to make mssql server work on travis

* Updated changelog and version

* Drop mariadb support  (#2681)

* Dropped support for mariasql

* ESLint

* Fixed docs to build again

* Fix knex.initialize() and adds test (#2477)

* Fix knex.initialize() and adds test

* Fix knex.initialize() and adds test

* Added test for reinitializing pool after destroy

* Fixed destroy / reinitialization test

* Fixed the fix

* Implement missing schema support for mssql dialect

* chore: Update dependencies. Remove estraverse (#2691)

* Update dependencies. Remove estraverse

* Fix compatibility with new Sinon

* Increase mssql timeout

* Normalized and validated driverNames of test db clients and fixed oracle test setup (#2692)

* Normalized and validated driverNames of test db clients and fixed oracle test setup

* Fixed failed queries from old query building tests which hadn't been ran in ages

* Allow selecting node version which is used to run oracledb docker tests

* Improved sql tester error messages

* Fixed rest of the oracledb tests

* Removed invalid flag from docker-compose

* Print mssql logs if initialization fails

* Fixed syntax error + final tests

* Added restart of failure for mssql DB initialization to try again if server was not ready

* Printout always mssql logs after container is started

* Fixed wait time printing after trying to connect

* Use npm run oracledb:test for testing oracle in travis

* Add Prettier (#2697)

* Add prettier
* Run files through prettier

* Istanbul -> NYC (#2700)

* istanbul -> nyc

* Update mocha

* Enforce code coverage (#2702)

* Enforce code coverage
* Update coverage numbers with current values

* version assignment on base class, copy on tx client, fix #2705

* Update changelog

* v0.15.1

* Added build step to test script. Fixes #2541 (#2712)

* Revert "Added build step to test script. Fixes #2541 (#2712)" (#2714)

This reverts commit 90ed8db.

* Allow oracle failures for now

* Fix issue with select(0) (#2711)

* Fix issue with select(0). Fixes #2658

* Refactor migrator (#2695)

* Refactor migrator

* Fix exports

* Fix ESLint

* Fix migrator

* Fix reference to config

* Split some more

* Fix table builder

* Fix argument order

* Merge branch 'master' of https://github.com/tgriesser/knex into feature/2690-support-multiple-migration-dirs

# Conflicts:
#	src/migrate/index.js
#	test/index.js
#	test/unit/migrate/migrator.js

* Fix #2715 (#2721)

* Fix #2715, explicitly set precision in datetime & timestamp
* Allow for precision in knex.fn.now, mysql time
* Add test for datetime with precision

* Bump changelog

* 0.15.2

* Fix issues with warnPromise when migration does not return a promise. Fixes #2725 (#2730)

* Add tests for multiple union arguments with callbacks and builders (#2749)

* Add tests for multiple union arguments

* Add some callback and raw tests to union unit tests

* Compile with before update so that bindings are put in correct order (#2733)

* Fix join using builder withSchema. (#2744)

* Use Datetime2 for MSSQL datetime + timestamp types (#2757)

* Use Datetime2 for MSSQL datetime + timestamp types

Datetime2 is now the recommended column type for new date work: https://docs.microsoft.com/en-us/sql/t-sql/data-types/datetime-transact-sql?view=sql-server-2017

* Add tests for MSSQL datetime2 changes

* General/document breaking change (#2774)

* Add changelog entry for a breaking change

* Improve entry

* Allow timestamp with timezone on mssql databases (#2724)

* Allow timestamp with timezone on mssql databases

* Change timestamp parameter to use object instead of boolean

* Update dependencies (#2772)

* Feature/2690: Multiple migration directories (#2735)

* Implement support for multiple migration directories

* Add tests for new functionality

* Fix tape tests

* Pass migration directory upwards

* Fix multiple directory support

* Fix bugs

* Rollback correctly

* Fix remaining tests

* Address comments

* #2758: Implement fail-fast logic for dialect resolution (#2776)

* Implement fail-fast logic for dialect resolution, clean-up code around.

* Remove method that was deprecated long time ago

* Address additional comments

* Try addressing comments

* Set client explicitly

* Fix compatibility with older Node versions

* Use columnize instead of wrap in using(). (#2713)

* Use columnize instead of wrap in using().

This is an attempt to fix #2136. Also added an integration test, couldn't find any existing.

* Exclude MSSQL from test as it does not support .using

* Change test to not use subquery

* Change test

* Introduced abstraction for getting migrations (#2775)

* Introduced abstraction for getting migrations

This would allow a webpack migration source which is compatible with bundling.

* Fixed migration validation with custom migration source

* Fixed issues after rebasing on muti directory PR

* Renamed parameter and fixed error message

* Addressed some PR comments

* Finished comment

* Moved filename extension filtering into fs-migrator

* Added test showing in memory custom migration source

* Cleaned up how to get config

* Fixed failing test

* Hopefully fix tests

* Fix Node.js 10 support in tests

* Test for correctly releasing cancel query connection
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment