Add insert benchmarking methods #27
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Benchmarking methods added for: * stream inserts using tuples of time, value data * stream inserts using pyarrow tables of timestamps, value columns * streamset inserts using a dict map of streamset stream uuids, and lists of tuples of time, value data * streamset inserts using a dict map of streamset stream uuids, and pyarrow tables of timestamps, values.
youngale-pingthings
added a commit
that referenced
this pull request
Jul 20, 2023
* Threadpool executor (#22) * Release v5.15.0 * update protobuf to v4.22.3 * Add threaded streamset calls Using concurrent.futures.ThreadPoolExecutor * Blacken code * Update for failing tests * Ignore flake8 as part of testing pytest-flake8 seems to have issues with the later versions of flake8 tholo/pytest-flake8#92 * Update .gitignore * Update ignore and remove extra print. * Remove idea folder (pycharm) --------- Co-authored-by: David Konigsberg <72822263+davidkonigsberg@users.noreply.github.com> Co-authored-by: Jeff Lin <42981468+jleifnf@users.noreply.github.com> * Threaded arrow (#23) * Release v5.15.0 * update protobuf to v4.22.3 * Add threaded streamset calls Using concurrent.futures.ThreadPoolExecutor * Blacken code * Update for failing tests * Ignore flake8 as part of testing pytest-flake8 seems to have issues with the later versions of flake8 tholo/pytest-flake8#92 * Update .gitignore * Update proto definitions. * Update endpoint to support arrow methods * Support arrow endpoints * Additional arrow updates * Update transformers, add polars conversion * Update .gitignore * Update ignore and remove extra print. * Remove idea folder (pycharm) * Update requirements.txt * Update btrdb/transformers.py * Update the way to check for arrow-enabled btrdb This has not been "turned on" yet though, since we dont know the version number this will be enabled for. The method is currently commented out, but can be re-enabled pretty easily. * Use IPC streams to send the arrow bytes for insert Instead of writing out feather files to an `io.BytesIO` stream and then sending the feather files over the wire, this creates a buffered outputstream and then sends that data back as bytes to btrdb. * Create arrow specific stream methods. * Update test conn object to support minor version * Update tests and migrate arrow code. * Arrow and standard streamset insert * Create basic arrow to dataframe transformer * Support multirawvalues, arrow transformers * Multivalue arrow queries, in progress * Update stream filter to properly filter for sampling frequency * Update arrow values queries for multivalues * Update param passing for sampling frequency * Update index passing, and ignore depth * benchmark raw values queries for arrow and current api * Add aligned windows and run func * Streamset read benchmarks (WIP) In addition: * update streamset.count to support the `precise` boolean flag. * Update mock return value for versionMajor * In progress validation of stream benchs --------- Co-authored-by: David Konigsberg <72822263+davidkonigsberg@users.noreply.github.com> Co-authored-by: Jeff Lin <42981468+jleifnf@users.noreply.github.com> * Add 3.10 python to the testing matrix (#21) * Add 3.10 python to the testing matrix * Fix yaml parsing * Update requirements to support 3.10 * Use pip-tools `pip-compile` cli tool to generate requirements.txt files from the updated pyproject.toml file * Include pyproject.toml with basic features to support proper extra deps * Support different ways to install btrdb from pip * `btrdb, btrdb[data], btrdb[all], btrdb[testing], btrdb[ray]` * Update transformers.py to build up a numpy array when the subarrays are not the same size (number of entries) * This converts the main array's dtype to `object` * tests still pass with this change * recompile the btrdb proto files with latest protobuf and grpc plugins * Create multiple requirements.txt files for easier updating in the future as well as a locked version with pinned dependencies * Ignore protoc generated flake errors * Update test requirements * Include pre-commit and setup. * Pre-commit lints. * Update pre-commit.yaml add staging to pre-commit checks * Fix missing logging import, rerun pre-commit (#24) * Add basic doc string to endpoint object (#25) * Update benchmark scripts. * Multistream read bench insert bench (#26) * Fix multistream endpoint bugs * The streamset was passing the incorrect params to the endpoint * The endpoint does not return a `version` in its response, just `stat` and `arrowBytes` Params have been updated and a NoneType is passed around to ignore the lack of version info, which lets us use the same logic for all bytes decoding. * Add multistream benchmark methods for timesnap and no timesnap. * Add insert benchmarking methods (#27) Benchmarking methods added for: * stream inserts using tuples of time, value data * stream inserts using pyarrow tables of timestamps, value columns * streamset inserts using a dict map of streamset stream uuids, and lists of tuples of time, value data * streamset inserts using a dict map of streamset stream uuids, and pyarrow tables of timestamps, values. * Fix arrow inserts (#28) * Add insert benchmarking methods Benchmarking methods added for: * stream inserts using tuples of time, value data * stream inserts using pyarrow tables of timestamps, value columns * streamset inserts using a dict map of streamset stream uuids, and lists of tuples of time, value data * streamset inserts using a dict map of streamset stream uuids, and pyarrow tables of timestamps, values. * Include nullable false in pyarrow schema inserts * This was the only difference in the schemas between go and python. * also using a bytesIO stream to act as the sink for the ipc bytes. * Start integration test suite * Add more streamset integration tests. * Add support for authenticated requests without encryption. * Optimize logging calls (#30) Previously, the debug logging in the api would create the f-strings no matter if logging.DEBUG was the current log level or not. This can impact the performance, especially for benchmarking. Now, a cached IS_DEBUG flag is created for the stream operations, and other locations, the logger.isEnabledFor boolean is checked. Note that in the stream.py, this same function call is only executed once, and the results are cached for the rest of the logic. * Add more arrow tests and minor refactoring. * More integration test cases * Restructure tests. * Mark new failing tests as expected failures for now. * Disable gzip compression, it is very slow. * Reenable test, server has been fixed. * Update pandas testing and fix flake8 issues (#31) * Update pandas testing and fix flake8 issues * Update stream logic for unpacking arrow tables, update integration tests. * add init.py for integration tests. * Add additional tests for arrow methods vs their old api counterparts. * Add tests for timesnap boundary conditions. (#32) * Add more integration tests. * Add additional integration tests, modify the name_callable ability of the arrow_values. * remove extraneous prints. * Include retry logic. * Update statpoint order in arrow, fix some bugs with the arrow methods. * Update testing to account for NaNs. * Update github action versions. * Update tests, add in a test for duplicate values. * Remove empty test, remove extraneous prints --------- Co-authored-by: andrewchambers <andrewchamberss@gmail.com> * Update docs for arrow (#35) * Update docs, add in final enhanced edits. * Only enable arrow-endpoints when version >= 5.30 (#36) Once we have a v5.30tag of the server with arrow/multistream, we can merge this and complete the ticket. * Update arrow notes, small doc changes. (#38) --------- Co-authored-by: Justin Gilmer <justin@pingthings.io> Co-authored-by: David Konigsberg <72822263+davidkonigsberg@users.noreply.github.com> Co-authored-by: Jeff Lin <42981468+jleifnf@users.noreply.github.com> Co-authored-by: Andrew Chambers <andrew@pingthings.io> Co-authored-by: andrewchambers <andrewchamberss@gmail.com>
justinGilmer
added a commit
that referenced
this pull request
Sep 25, 2023
* Threadpool executor (#22) * Release v5.15.0 * update protobuf to v4.22.3 * Add threaded streamset calls Using concurrent.futures.ThreadPoolExecutor * Blacken code * Update for failing tests * Ignore flake8 as part of testing pytest-flake8 seems to have issues with the later versions of flake8 tholo/pytest-flake8#92 * Update .gitignore * Update ignore and remove extra print. * Remove idea folder (pycharm) --------- Co-authored-by: David Konigsberg <72822263+davidkonigsberg@users.noreply.github.com> Co-authored-by: Jeff Lin <42981468+jleifnf@users.noreply.github.com> * Threaded arrow (#23) * Release v5.15.0 * update protobuf to v4.22.3 * Add threaded streamset calls Using concurrent.futures.ThreadPoolExecutor * Blacken code * Update for failing tests * Ignore flake8 as part of testing pytest-flake8 seems to have issues with the later versions of flake8 tholo/pytest-flake8#92 * Update .gitignore * Update proto definitions. * Update endpoint to support arrow methods * Support arrow endpoints * Additional arrow updates * Update transformers, add polars conversion * Update .gitignore * Update ignore and remove extra print. * Remove idea folder (pycharm) * Update requirements.txt * Update btrdb/transformers.py * Update the way to check for arrow-enabled btrdb This has not been "turned on" yet though, since we dont know the version number this will be enabled for. The method is currently commented out, but can be re-enabled pretty easily. * Use IPC streams to send the arrow bytes for insert Instead of writing out feather files to an `io.BytesIO` stream and then sending the feather files over the wire, this creates a buffered outputstream and then sends that data back as bytes to btrdb. * Create arrow specific stream methods. * Update test conn object to support minor version * Update tests and migrate arrow code. * Arrow and standard streamset insert * Create basic arrow to dataframe transformer * Support multirawvalues, arrow transformers * Multivalue arrow queries, in progress * Update stream filter to properly filter for sampling frequency * Update arrow values queries for multivalues * Update param passing for sampling frequency * Update index passing, and ignore depth * benchmark raw values queries for arrow and current api * Add aligned windows and run func * Streamset read benchmarks (WIP) In addition: * update streamset.count to support the `precise` boolean flag. * Update mock return value for versionMajor * In progress validation of stream benchs --------- Co-authored-by: David Konigsberg <72822263+davidkonigsberg@users.noreply.github.com> Co-authored-by: Jeff Lin <42981468+jleifnf@users.noreply.github.com> * Add 3.10 python to the testing matrix (#21) * Add 3.10 python to the testing matrix * Fix yaml parsing * Update requirements to support 3.10 * Use pip-tools `pip-compile` cli tool to generate requirements.txt files from the updated pyproject.toml file * Include pyproject.toml with basic features to support proper extra deps * Support different ways to install btrdb from pip * `btrdb, btrdb[data], btrdb[all], btrdb[testing], btrdb[ray]` * Update transformers.py to build up a numpy array when the subarrays are not the same size (number of entries) * This converts the main array's dtype to `object` * tests still pass with this change * recompile the btrdb proto files with latest protobuf and grpc plugins * Create multiple requirements.txt files for easier updating in the future as well as a locked version with pinned dependencies * Ignore protoc generated flake errors * Update test requirements * Include pre-commit and setup. * Pre-commit lints. * Update pre-commit.yaml add staging to pre-commit checks * Fix missing logging import, rerun pre-commit (#24) * Add basic doc string to endpoint object (#25) * Update benchmark scripts. * Multistream read bench insert bench (#26) * Fix multistream endpoint bugs * The streamset was passing the incorrect params to the endpoint * The endpoint does not return a `version` in its response, just `stat` and `arrowBytes` Params have been updated and a NoneType is passed around to ignore the lack of version info, which lets us use the same logic for all bytes decoding. * Add multistream benchmark methods for timesnap and no timesnap. * Add insert benchmarking methods (#27) Benchmarking methods added for: * stream inserts using tuples of time, value data * stream inserts using pyarrow tables of timestamps, value columns * streamset inserts using a dict map of streamset stream uuids, and lists of tuples of time, value data * streamset inserts using a dict map of streamset stream uuids, and pyarrow tables of timestamps, values. * Fix arrow inserts (#28) * Add insert benchmarking methods Benchmarking methods added for: * stream inserts using tuples of time, value data * stream inserts using pyarrow tables of timestamps, value columns * streamset inserts using a dict map of streamset stream uuids, and lists of tuples of time, value data * streamset inserts using a dict map of streamset stream uuids, and pyarrow tables of timestamps, values. * Include nullable false in pyarrow schema inserts * This was the only difference in the schemas between go and python. * also using a bytesIO stream to act as the sink for the ipc bytes. * Start integration test suite * Add more streamset integration tests. * Add support for authenticated requests without encryption. * Optimize logging calls (#30) Previously, the debug logging in the api would create the f-strings no matter if logging.DEBUG was the current log level or not. This can impact the performance, especially for benchmarking. Now, a cached IS_DEBUG flag is created for the stream operations, and other locations, the logger.isEnabledFor boolean is checked. Note that in the stream.py, this same function call is only executed once, and the results are cached for the rest of the logic. * Add more arrow tests and minor refactoring. * More integration test cases * Restructure tests. * Mark new failing tests as expected failures for now. * Disable gzip compression, it is very slow. * Reenable test, server has been fixed. * Update pandas testing and fix flake8 issues (#31) * Update pandas testing and fix flake8 issues * Update stream logic for unpacking arrow tables, update integration tests. * add init.py for integration tests. * Add additional tests for arrow methods vs their old api counterparts. * Add tests for timesnap boundary conditions. (#32) * Add more integration tests. * Add additional integration tests, modify the name_callable ability of the arrow_values. * remove extraneous prints. * Include retry logic. * Update statpoint order in arrow, fix some bugs with the arrow methods. * Update testing to account for NaNs. * Update github action versions. * Update tests, add in a test for duplicate values. * Remove empty test, remove extraneous prints --------- Co-authored-by: andrewchambers <andrewchamberss@gmail.com> * Update docs for arrow (#35) * Update docs, add in final enhanced edits. * Only enable arrow-endpoints when version >= 5.30 (#36) Once we have a v5.30tag of the server with arrow/multistream, we can merge this and complete the ticket. * Update arrow notes, small doc changes. (#38) * fix: patch up stream object type and other bugs (#33) * fix: patch up stream object type and other bugs * fix: resolve depth errors in stream window * fix: resolve remaining test warnings * fix: resolve test imports * chore: add pre-commit install to readme * Update staging branch with latest `master` changes (#52) --------- Co-authored-by: David Konigsberg <72822263+davidkonigsberg@users.noreply.github.com> Co-authored-by: Jeff Lin <42981468+jleifnf@users.noreply.github.com> Co-authored-by: Andrew Chambers <andrew@pingthings.io> Co-authored-by: andrewchambers <andrewchamberss@gmail.com> Co-authored-by: Taite Nazifi <135669716+pingt8@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Benchmarking methods added for:
stream inserts using tuples of time, value data
stream inserts using pyarrow tables of timestamps, value columns
streamset inserts using a dict map of streamset stream uuids, and lists of tuples of time, value data
streamset inserts using a dict map of streamset stream uuids, and pyarrow tables of timestamps, values.