CHANGES.rst

2024-10-07 0.27.2

• Changed run-crate <version> to download the x64 version of a tarball under a arm64 Mac because there are no mac arm64 distributions for CrateDB. The x64 version works via emulation layer.

2024-09-09 0.27.1

• Fixed a regression in 0.27.0 that could cause HTTP sessions to remain open, resulting in stuck operations.

2024-09-09 0.27.0

• Added support for setting up session settings per connection for spec files.
• Fixed an issue that caused CrateDB process leaks if cr8 run-crate was killed via SIGTERM.

2024-03-25 0.26.1

• Fixed a regression that could lead to errors using cr8 run-crate when it tries to extract the CrateDB tarball.

2024-03-20 0.26.0

• Added support for python 3.12
• Fixed a permission denied error that could occur using cr8 run-crate with a branch version specification.

2023-10-24 0.25.0

• Added compatibility for argh>=0.30.1. Due to this several of the main entry point functions have now mandatory keyword arguments and no longer support positional arguments. For CLI use nothing changes, but for API consumers this is a breaking change.

2023-10-12 0.24.0

• Fixed an issue that prevented cr8 run-crate path/to/source from working if used as part of git bisect run.
• Added handling for maven as build system when using cr8 run-crate with a source path or branch:<name>.
• Changed cr8 insert-from-sql to quote the column names used in the insert statements to make them case sensitive.

2023-07-10 0.23.0

• Use tomllib if available (Python 3.11+) in favour of toml. The tomllib parser seems more resilient when parsing multi-line strings.
• Fixed a column lookup error in insert-fake-data that happened when using a mapping file.
• Increased the timeout for CrateNode.stop() to give the CrateDB shutdown procedure more time to shutdown gracefully.

2023-01-16 0.22.0

• Added support for HTTP(s) URLs as source to load_data of spec files.
• Added a --re-name option to the run-spec command which allows to only run queries where the name property matches the given regex.
• Added a warmup option to queries in spec files
• Dropped Python 3.6 Support

2022-02-08 0.21.1

• Changed cr8 run-crate to implicitly create the cache folder if it is missing instead of raising an error.
• Improved the performance of the insert-fake-data ``auto_inc provider for id columns of type integer.

2022-01-26 0.21.0

• Added a setup.data_cmds directive to allow data generation via external commands within spec files. The command must output one JSON object per line.
• Added a data provider for the bit data type.

2020-09-23 0.20.1

• Added asyncpg as non-optional dependency, as it is required by insert-from-sql.

2020-09-22 0.20.0

• Added a new insert-from-sql command that can be used to insert data from another CrateDB or PostgreSQL server.
• Extended insert-json to also support a list of JSON objects within a single line as input format. (Something like [{"name": "n1"}, {"name": "n2"}])

2020-07-15 0.19.3

• run-crate now uses a system specific CrateDB tarball if available. This is required for the platform dependent builds starting with CrateDB 4.2.

2020-06-10 0.19.2

• Extended the faker version constraint to >= 4, < 5, this should make it easier to use cr8 as a dependency in projects where another dependency uses faker as well.

2020-05-14 0.19.1

• Extended run_crate to be forward compatible with an upcoming change in the log format in CrateDB.
• Fixed the handling of the ?verify_ssl=false parameter in --hosts

2020-05-02 0.19.0

• Spec files now support a name property in the queries section to make it easier to analyse the results.
• run-spec can now also be used with PostgreSQL if using asyncpg.
• reindex now skips blob tables and quotes schema and table names, so it should work in more cases.

2020-03-27 0.18.0

• timeit can now be used with PostgreSQL and Cockroach if using asyncpg.
• insert-fake-data now uses uuid4 automatically for columns of type text. This was already the case pre CrateDB 4.0 for columns of type string.
• run-spec now implicitly triggers an ANALYZE after the setup.

2019-08-20 0.17.0

• Added a new reindex command which will go through all tables on a cluster which need to be re-indexed to be compatible with CrateDB major_version+1. It will one by one create a copy of a table and then remove the old table, replacing it with the copy.
• Added a new infile argument to the insert-json command. This should make certain scripting scenarios easier.

2019-07-01 0.16.0

• Added type mapping for the timestamp with time zone and timestamp without time zone data types so that insert-fake-data can generate data for those columns by default.
• run-crate now supports arbitrary branch builds. E.g. run-crate branch:master
• Removed the LineBuffer from run_crate and added support for callables instead.

2019-04-04 0.15.1

• Fixed a compatibility issue with Python 3.6
• Adapted some queries and type mappings to be compatible with CrateDB 4.0
• Corrected the minimum CrateDB version that can be run with Java 11.
• Added a version arg to CrateNode which can be used to overrule the auto-detection.

2019-02-28 0.15.0

• run-crate will now try to choose a different JAVA_HOME if the given version of CrateDB can't be run with the default JAVA_HOME. This behavior can be disabled with --disable-java-magic. The behavior of the CrateNode API is unchanged and by default won't try to change the JAVA_HOME.
• run-crate will now avoid re-building branches from source if there aren't any new commits.

2019-02-14 0.14.2

• run-crate <release_branch> will now make sure that the sources are updated to avoid stale builds.

2019-02-05 0.14.1

• Made run-crate where the argument is a path to a repository forward compatible with upcoming build changes.

2018-10-08 0.14.0

• SSL validation can no be disabled by including verify_ssl=False in the hosts URI.
• insert-fake-data now generates timestamps differently so it works with asyncpg
• JSON output is no longer pretty printed by default. Use jq or python -mjson.tool to do so.
• Release branches can now be used as argument to run-crate. (Something like run-crate 3.1. This will result in a source checkout and the tarball will be built locally.)
• Improved the error message when connecting via HTTP and running into a 401.

2018-07-04 0.13.0

• Added experimental postgres protocol support. It's available if the optional asyncpg dependency is installed.
• Bumped the aiohttp dependency for Python 3.7 support.
• The load_data directive in spec files now can read gzipped files.
• Cached local tarballs are now checksummed to avoid re-using a stale tarball from cache.
• Samples and stdev are now included in the runtime_stats output if only 1 sample is available.

2018-06-08 0.12.1

• run-track now exists with an error code if any statement failed.
• The statements printed during run-spec are no longer trimmed.
• Version wildcards like 2.3.x now work correctly for digits greater than 9.

2018-05-24 0.12.0

• Fixed an issue that caused insert-fake-data to fail with a TypeError.
• spec files written in python can now use generators for statements or arguments.
• Improved the fake data generation for insert-fake-data. It now works for arrays and objects (although they'll simply be empty)
• Added a duration option to spec files and timeit
• Added a sample-mode option to control how many samples will be kept for the results.
• Improved some error handling and error reporting

2018-02-04 0.11.1

• The disk.watermark settings are no longer set by default by run-crate in order to be compatible with CrateDB 3.0
• run-crate will now exit with a failure if process chaining is used and one of the chained processes failed.
• stop() on CrateNode now resets certain attributes correctly, so that start() doesn't fail with connection errors.

2017-11-05 0.11.0

• Added a new default provider for columns of type BYTE.
• Added a new default provider for columns of type GEO_SHAPE, which provides a POLYGON WKT string.
• Dropped support for Python 3.5
• run-crate now correctly supports settings using unicode characters.
• run-crate will now remove old tarballs from the cache folder after a while.
• run-crate should now fail faster if an invalid setting is used.
• run-crate now supports arbitrary command chaining using -- @cmdname If command chaining is used, run-crate will terminate after all commands have been run.
• run-crate should now work correctly if CrateDB is bound to a IPv6 address.

2017-09-12 0.10.2

• insert-fake-data should no longer generate the same values using the uuid4 provider. The amount of duplicate values generated using other providers should be reduced as well.

2017-08-04 0.10.1

• run-crate now works again with latest-nightly. It ran into a timeout as it couldn't parse the HTTP address from the log due to a format change.

2017-06-30 0.10.0

• Added a new --fail-if argument to timeit and run-spec.

• Added support for sub-command chaining using --. This is especially useful if the first command is run-crate. Together with --fail-if this can be used with git bisect to determine the first commit that introduced a performance regression. An example:

  cr8 run-crate /path/to/crate/src

  -- timeit -s "select... " --hosts '{node.http_url}' --fail-if "{runtime_stats.mean} > 1.34"

insert-fake-data

• Added a default provider for columns of type short.

run-crate

• Pass LANG environment variable to crate subprocess. This fixes encoding issues when passing unicode characters as CrateDB setting values.
• It's now possible to launch SSL enabled nodes. Before run-crate would run into a timeout.
• The version identifier can now include x as wildcard. For example, use: run-crate 2.0.x to run the latest hotfix version in the 2.0 series.
• Added support for building and running crate from a source tree.
• Environment variables set using --env can now contain = signs.

2017-05-14 0.9.3

• insert-fake-data: Increased the default value range for columns of type integer or long.
• Updated aiohttp to version 2
• insert-fake-data: The schema and table name is now quoted to allow using reserved keywords as schema or table name.

2017-02-11 0.9.2

• Values of type Decimal or datetime can now be serialized. This fixes an issue that could cause insert-fake-data to not work with schemas that contained columns of type double. It also allows track files written in python to use Decimal or datetime objects as arguments.
• If python-argcomplete is installed and registered that should now be picked up to enable tab-completion in bash.
• Fixed an issue that caused warnings with newer aiohttp versions.
• Adapted run-crate to handle upcoming breaking changes. It's now able to launch tarballs of CrateDB 1.1 and 1.2. snapshots.

2017-01-03 0.9.1

• Fixed an issue that caused failures on Windows
• timeit can now receive multi-line statements via stdin

2016-12-13 0.9.0

Breaking

• Changed the default output format to text. In addition, the values of --output-format were renamed from full and short to json and text.

Miscellaneous improvements

• Added a --keep-data option to run-crate. If this is set the data folder isn't removed if the process is stopped.
• The version argument of run-crate can now also be a fs path to a CrateDB tarball.
• Various error handling and Ctrl+c improvements.
• Added --logfile-info and --logfile-result options to run-spec and run-track.

2016-11-12 0.8.1

• Fixed a regression that caused run-spec to save results into hosts instead of result-hosts.

2016-11-10 0.8.0

insert-fake-data improvements

• Multiple cores are now utilized better for fake data generation.
• Adopted internal queries to be compatible with Crate versions > 0.57.
• insert-fake-data will now insert the accurate number of rows specified instead of rounding to the nearest bulk size.

Miscellaneous

• run-crate latest-stable now correctly launches the latest released stable version of Crate. It incorrectly retrieved the version of the latest Java client release.
• run-crate now outputs the postgres port if found in the logs.
• Added a --action argument to run-spec which can be used to only run a subset of a spec file.
• Extended the track-file format to allow re-using a setup across multiple spec files.
• Added a --version option. Best feature ever.
• Changed the --help output formatting so it's easier to read.

2016-10-14 0.7.0

Breaking (but only a little)

• The run-track subcommand now does not fail any more if a single spec file of the track fails. To achieve the same behaviour as before, you can use the --failfast command line option.

Improvements

• Added a --output-fmt option to most commands. This option can be used to get a succinct output.
• Added a new auto_inc fake data provider for insert-fake-data. This provider may be a bit slow. This is due to the fact that the fake data generation utilizes multiple processes and this provider requires synchronization. But it's still awesome.
• Spec files now support a min_version setting. This can be used to skip certain queries if the server doesn't meet the min_version requirement.
• Improved the error handling a bit.
• Statements and arguments in spec files can now be defines as callables.
• Added meta object column to results table. It's now possible to add a name to the spec so the benchmark results can easily be identified by this spec label.
• Added Crate build date column to version_info in benchmark result table.
• timeit now shows a progress bar.
• The --setting and -env options of run-crate are now repeatable.

Fixes

• Fixed an issue with the num-records option of insert-fake-data. It didn't work correctly if the number of records specified was smaller than the bulk size.
• Fixed some issues with the way Crate is launched using run-track. If Crate produced a lot of logging output it could get stuck.

2016-07-04 0.6.0

Breaking

• hosts and table is now always a named argument. This affects timeit, insert-json, insert-blob and insert-fake-data

Features 🍒

run-track

Added a new run-track command. This command can be used to execute track files. A track file is a file in TOML format containing a matrix definition of Crate versions, Crate configurations and spec files.

The command will run each listed Crate version with each configuration and run all listed spec files against it.

Other improvements

• Added a new run-crate command.
• Added a fake-data provider for geo_point columns.
• Improved the --help output of most commands.
• Run-spec output is now proper JSON
• Spec files can be written in python
• args and bulk_args can now be specified in toml spec files.

Fixes 💩

• runtime_stats['n'] is no longer capped to 1000
• insert-json now ignores empty lines instead of causing an error.

2016-06-09 0.5.0

Breaking 💔

Pretty much everything:

• Renamed blob upload to insert-blob
• Renamed json2insert to insert-json
• Renamed fill-table to insert-fake-data
• Removed find-perf-regressions

New & shiny features ✨

run-spec

Added a new command which can be used to "run" spec files. Spec files are either .json or .toml files which contain setup, queries and tear-down directives. A minimal example:

[setup]
statement_files = ["sql/create_countries.sql"]

    [[setup.data_files]]
    target = "countries"
    source = "data/countries.json" # paths are relative to the spec file

[[queries]]
statement = "select count(*) from countries"
iterations = 1000

[teardown]
statements = ["drop table countries"]

run-spec will execute the given specification and output runtime statistics. The result can also directly be inserted into a Crate cluster.

insert-fake-data & insert-json

• Column names are now quoted in the insert statement

insert-fake-data

• No longer tries to generate data for generated columns
• Speed improvements
• Added default provider mappings for columns of type float, double and ip

insert-json

• Prints runtime stats after the inserts are finished

timeit

• Added a histogram and percentiles to the runtime statistics that are printed
• Added a concurrency option

2016-05-19 0.4.0

• Python 3.4 support has been dropped.
• Subcommands that take numbers as arguments now support python literal notation. So something like 1e3 can be used.

Features

fill-table

• Consumes less memory and is faster since it no longer generates all data upfront but starts inserting as soon as possible.
• Added a concurrency option to control how many requests to make in parallel (at most).
• Columns of type long are automatically mapped to the random_int provider.

json2insert

• json2insert can now be used to bulk insert JSON files. The following input formats are supported:

  1 JSON object per line
  
      {"name": "n1"}
      {"name": "n2"}
  
  Or 1 JSON object:
  
      {
          "name": "n1"
      }
  
  Or a list of JSON objects:
  
      [
          {"name": "n1"},
          {"name": "n2"},
      ]
  

  The input must be fed into stdin.

• The --bulk-size and --concurrency options have been added.