v0.17.0

Breaking changes

• UAST UDFs now returns data in a specific format, in a BLOB column.

• You must regenerate indexes to use this version. pilosalib index driver is now called pilosa, and the old pilosa driver does not exists anymore.

Bugfixing

• *: pushdown projections properly in the files table
• remove oniguruma support from prebuilt binaries
• gitbase: close repositories in squash table
• pilosalib: fix index usage with parallelism
• sql/plan: make sure ResolvedTable closes row iterators
• vendor: upgrade go-git, fix closed file in pack iter
• gitbase: remove warnings on symbolic references
• function: fix race condition getting UASTs
• sql/analyzer: check non-parallelizable nodes under QueryProcess
• sql/analyzer: execute parallelize as last rule
• rule: do not error when parts of joins are not squashable
• gitbase: close siva FS after use
• If deferredColumns is an alias then resolve it.
• Fix get/set from global table.
• build: disable static linking and oniguruma
• rule: fix unability to check conditions of natural joins

New features

• A new way to serialize UASTs
• uast_children new UDF
• uast_extract new UDF
• *: implement show full processlist
• *: implement session configuration
• sql/index: remove old pilosa driver and rename pilosalib to pilosa
• gitbase: implement sql.PartitionCounter in all tables
• *: correctly handle process management using processlist
• sql/plan: make create index sync by default
• sql/analyzer: resolve aggregate aliases in groupby grouping
• Upgrade mysql and vitess.
• build: only tag latest in docker when version is not prerelease.
• build: make static build with oniguruma enabled
• internal/function: modify uast_extract to return a flattened array
• sql: make table resolution case insensitive
• Support ON/OFF for SET autocommit
• Update go-mysql-server to do not force timeouts
• Add timeout on the configuration

Performance improvements

• Update pilosa to the latest version
• Add partitions to process several repositories at the same time
• Add UAST cache
• Add language cache
• Compile gitbase using oniguruma instead regular regex expressions to make LANGUAGE function faster
• One cache for repository pool.
• gitbase: use KeepDescriptors go-git option
• *: make SquashedTable not implement sql.Node

Documentation

• Make some snippets copyable
• docs: fix how to install oniguruma on linux
• docs: fix uast function descriptions to reflect the new uast serialization
• docs: explain how uast and uast_mode functions serialize data
• docs: add GITBASE_CACHESIZE_MB to the documentation
• docs: add note about oniguruma support in static binary
• docs: guide on how to optimize queries
• docs: add code example on how to manage serialized uast data
• docs: fix optional dependencies
• docs: Add instructions how to run in docker with bblfsh
• docs: improve uast functions documentation