Add parameter comment in timestamp API
I add parameter comment in timestamp API. I commented to make a input as string. Please look on. Author: kwangin <inylove82@gmail.com> Closes #1064 from djKooks/ts-comment and squashes the following commits: 82235d5 [kwangin] Update wrong typo & return type 8372333 [kwangin] Add parameter comment in timestamp API
ENH/PERF: Speed up pandas groupby
This prevents individual group evaluation for things that do not need it Author: Phillip Cloud <cpcloud@gmail.com> Closes #1067 from cpcloud/fix-pandas-gb-perf and squashes the following commits: adf2963 [Phillip Cloud] Fix again 30b8f44 [Phillip Cloud] Fix benchmark f32ef1e [Phillip Cloud] Add benchmark 185f68f [Phillip Cloud] ENH/PERF: Generic binary operations on SeriesGroupBy
ENH: Implement comparison for string vs date and timestamp types
PERF: Fix execution perf of groupby and selection
Author: Phillip Cloud <cpcloud@gmail.com> Closes #1073 from cpcloud/pandas-perf and squashes the following commits: 26bfaca [Phillip Cloud] Do not fail if benchmarks are off b5011ed [Phillip Cloud] Of course we need to fetch 82927cc [Phillip Cloud] Make sure we are checking against upstream ibis and not ourselves 31e9373 [Phillip Cloud] PERF: Fix execution perf of groupby and selection
BUG: Memoization should include expression name where available
Also: - [x] fix docker builds - [ ] ~upgrade docker image to `cpcloud86/impala:java8`~ - [x] remove most bare `except:`s - [x] fix a broken doc test Author: Phillip Cloud <cpcloud@gmail.com> Closes #1080 from cpcloud/fix-named-impala-exprs and squashes the following commits: a27c334 [Phillip Cloud] Move back to java7 f770940 [Phillip Cloud] CI: Try skipping hive things f8cc86b [Phillip Cloud] CI: Docker image envar e920040 [Phillip Cloud] BLD: Upgrade docker image to cpcloud86/impala:java 8 8f8d5c2 [Phillip Cloud] CLN: Remove bare excepts where possible aafe1c3 [Phillip Cloud] ENH: More generic get_name 52a12e6 [Phillip Cloud] BUG: Memoization should include expression name where available 72d9da3 [Phillip Cloud] TST: Add failing test
ENH: Enable unary math operations for pandas, sqlite
Implement decimal for pandas Add SQLite unary ops Fix operations in postgres that require numeric Author: Phillip Cloud <cpcloud@gmail.com> Closes #1071 from cpcloud/unary-ops and squashes the following commits: 57ff2b1 [Phillip Cloud] BUG: Pass args and kwargs d94b0c6 [Phillip Cloud] REF: Factor vectorize object function 0d863d1 [Phillip Cloud] ENH: Enable unary math operations for pandas, sqlite
REF: Refactor the postgres test suite
Author: Phillip Cloud <cpcloud@gmail.com> Closes #1078 from cpcloud/refac-pg-tests and squashes the following commits: 5c46473 [Phillip Cloud] Do not remove our cache 2bb70bc [Phillip Cloud] Add caching back d6118ae [Phillip Cloud] Do not check column type on python 2 37c7685 [Phillip Cloud] CLN/CI/BLD: Remove the cache 8f00958 [Phillip Cloud] BUG: Make sure we do not have sqlalchemy subclasses as types 66483a0 [Phillip Cloud] Fix client tests 9cec1a5 [Phillip Cloud] Move postgres fixtures to conftest eb94061 [Phillip Cloud] Move common to conftest 58e281a [Phillip Cloud] REF: Refactor the postgres test suite
CI: Add hive metastore information back to CI
Author: Phillip Cloud <cpcloud@gmail.com> Closes #1081 from cpcloud/add-hive-metastore-listing-back-to-build and squashes the following commits: 2af5908 [Phillip Cloud] CI: Remove unnecessary echoing 4de0e4d [Phillip Cloud] CI: Proper path fad1884 [Phillip Cloud] CI: Print out more information when saving c4415bb [Phillip Cloud] CI: Remove cache if changing docker image 0ec2571 [Phillip Cloud] CI: Add information back to impala setup
skip testing on not-installed impala packages skip testing sqlite on non-existant testing db add useful Makefile directives Author: Jeff Reback <jeff@reback.net> Closes #1089 from jreback/testing and squashes the following commits: a2ed00b [Jeff Reback] skip testing on not-installed impala packages skip testing sqlite on non-existant testing db add useful Makefile directives
Default to literal for as_value_expr calls in argument validation
Author: Phillip Cloud <cpcloud@gmail.com> Closes #1095 from cpcloud/fix-datatype-comparison and squashes the following commits: c313c75 [Phillip Cloud] Custom as value expr 5d6ceb4 [Phillip Cloud] Remove explicit parent class calls 48e2b2b [Phillip Cloud] Fix equality to short circuit on Any type c7fdb4b [Phillip Cloud] Add array rule test
ENH: Move pandas fixtures to conftest
Author: Phillip Cloud <cpcloud@gmail.com> Closes #1102 from cpcloud/pandas-conftest and squashes the following commits: b2cb1d8 [Phillip Cloud] Add pandas marker d66f631 [Phillip Cloud] Fix python 2 imports 383a0ab [Phillip Cloud] Move tz fixture to the only place it is used ba450b7 [Phillip Cloud] Fix test to only use the tz fixture if necessary 56cdfc4 [Phillip Cloud] Move fixtures to conftest db45938 [Phillip Cloud] Add conftest
ENH: Pass ops in scope instead of exprs
Also fixes the `now` failures. Closes #1106. Author: Phillip Cloud <cpcloud@gmail.com> Closes #1109 from cpcloud/convert-expr-to-op-inscope and squashes the following commits: f50255b [Phillip Cloud] BUG: Execute with con not execute method 2421c26 [Phillip Cloud] BUG/TST: Construct and execute close in time to avoid pathlogical failures across the hour eca434a [Phillip Cloud] CLN: Remove unused code 169cc02 [Phillip Cloud] CLN: Remove execute_first for now 26d3c86 [Phillip Cloud] ENH: Pass ops in scope instead of exprs
ENH/CI: Migrate builds to CircleCI 2.0
This ~~cuts our build times in half~~ keeps our build times around the same (on average a little shorter), even though we're now running conda build in addition to all the previous steps. Closes #1057 Closes #1115 Closes #1116 Author: Phillip Cloud <cpcloud@gmail.com> Closes #1113 from cpcloud/circle2 and squashes the following commits: 91ca661 [Phillip Cloud] ENH/CI: Migrate builds to CircleCI 2.0
ENH/CI: Fix data loading in CI
* testing postgres on windows (w00t!) * `ci/datamgr.py` is a cross platform way to download the test data + create sqlite and postgres databases ready for local *or* ci testing. * remove unused code from `test_data_admin.py` (this now only uploads data and built UDFs to impala) Author: Phillip Cloud <cpcloud@gmail.com> Closes #1117 from cpcloud/cleanup-data-loading and squashes the following commits: 19c0aa1 [Phillip Cloud] ENH: Add datamgr script, clean up data loading
BUG: timestamp('America/Los_Angles') was not working as a substitute …
…for dt.Timestamp('America/Los_Angles')
because of lowercasing of the timezone in type validation
Author: Jeff Reback <jeff@reback.net>
Closes #1134 from jreback/tz and squashes the following commits:
fcce14b [Jeff Reback] BUG: timestamp('America/Los_Angles') was not working as a substitute for dt.Timestamp('America/Los_Angles')ENH: Pass scope around to execute_first
This is a small fix to make sure that scope gets passed around to every dispatched function so that those functions can take advantage of bound parameters. Author: Phillip Cloud <cpcloud@gmail.com> Closes #1133 from cpcloud/pass-scope-to-execute-first and squashes the following commits: c18abdd [Phillip Cloud] Merge scope in execute_first on window ops c7c829c [Phillip Cloud] BUG: Mark pytest.mark.pandas for split up tests eacac4b [Phillip Cloud] ENH: Pass scope around to execute_first
DISP: Show a more informative table type name for pandas client tables
ENH: Add a data_preload step in pandas backend
This PR adds a `data_preload` step to the execution pipeline for the Pandas backend. The motivation for this is to allow clients to perform operations on `(Node, ConcreteData)` pairs (including a `scope` argument) *before* the execution for a particular operation starts. Custom data source execution is the main use case here. As a motivating example, consider a custom data source object that can be turned into a DataFrame. We want to be able to operate on our custom data source, but we don't want to redefine every operation for this data source, or it doesn't make sense to define operations on this custom object directly. `data_preload` gives the ability to call a function on data pieces that are in scope before any execution happens. In the example above, `data_preload` would turn the `ConcreteData` object into a `pandas.DataFrame`. The default behavior is no-op. This also sets the stage for multi-client execution. More on that to come. Author: Phillip Cloud <cpcloud@gmail.com> Closes #1142 from cpcloud/data-preload and squashes the following commits: 11124fd [Phillip Cloud] ENH: Add a data_preload step in pandas backend
Add convenience specification for tz casting in time().between(....)