1.0.0 RC 11

Release Candidate 11

Fixes

• Incorrect data for queries spanning multiple tags with identical fields - #19
• Correctly match all requested tags when searching metric paths for multi tag queries on same measurement

1.0.0 RC 10

Release Candidate 10

Note that this and subsequent releases require pip version >= 6.0.0 if installing via pip

Fixes

• Incorrect values returned for multiple tag queries on same measurement - #14
• Incorrect values returned for multiple path queries without template configuration - #18

Changes

• Saved index reverted back to uncompressed to improve save performance and reduce memory requirements

1.0.0 RC9

Release Candidate 9.

Fixes:

• Multiple request target paths with unique queries not returning data for all paths - #13
• Template filters not being applied on query generation
• Template pattern being incorrectly matched on a subset of a metric path
• Template filtering now based on string sub-path searches as per InfluxDB
• Wrong data assigned to metric path when querying multiple differing paths from differently matched templates

1.0.0 RC 8

Release Candidate 8.

This RC contains many changes since RC7 to improve index performance and lower memory usage.

Memory usage has decreased 4-5 times across the board while index building performance has increased 4 times on CPython when C extensions are used.

Native C extensions will now be used when built for the platform for index and template functionality to further increase performance on CPython interpreters. PyPy is still recommended as it performs best all around and does not require native extensions.

Barring bug fixes in subsequent RCs, a 1.0 release will be up next.

Changes

• Finder will now not log anything by default unless explicitly enabled in configuration
• Index and template functionality now optionally available as C extensions to improve performance
• Re-worked index implementation to lower memory requirements and improve performance

1.0.0 beta 21

Fixes:

• Index C extension operations blocking webapp - #12
• Wait for an in-progress index save to finish before starting a new index build/save

1.0.0 beta 20

Fixes:

• Regression on index queries with unicode strings
• Avoid re-saving to disk, at startup, an index that was loaded from disk

1.0.0 beta 19

Changes:

• C extension for critical index building parts of template functionality
• Performance improvements for index and template extensions

1.0.0 beta 18

Changes:

• Index query performance optimisations for index C extension

Fixes:

• Regression on loading index from file
• Index saving blocking main Python thread for duration of file write
• Disable wheel building on Travis CI so that PyPi deployment does not fail as we now include C extension

1.0.0 beta 17

Changes:

• Index tree C extension implementation to improve performance
• Index tree optimisations to reduce index memory footprint - resolves #10
• Metric names stored as bytes on index to reduce memory footprint and decoded to UTF-8 encoded strings on retrieval.
• Saved indexes from earlier releases are not compatible with this release and will load only partial data - best to wipe any saved indexes or wait for next rebuild.

On Py2 (only) unicode metric names is a change in behaviour where previously names would remain as byte strings if inserted as byte strings. In practice this is only an implementation detail as all strings in InfluxDB are UTF-8 encoded.

1.0.0 beta 16

Fixes:

• Fix index load and test post re-factor

Changes:

• Further index building performance optimisations
• Updated setup.py requirements