v2.6.0

Arrow v2.6.0

Diff since v2.5.2

Closed issues:

• Support Tables.partitions when reading "arrow file" format in addition to "stream" format (#293)
• Make recursive iteration to get dictionaries more defensive for interop (#375)
• Error/Segfault when writing many partitions (#396)
• Vector{UInt8} mis-represented when writing to disk (#411)
• CI doesn't test with multiple threads (#426)
• Malformed file by Arrow.write on a IOStream created with open(filename, "w") (#432)
• Unhandled sentinel value for len in compression causes invalid Array dimensions (#435)

Merged pull requests:

• Get dictionaries of children only when field.children not nothing (#382) (@okartal)
• fix Base.eltype methods and functions that take Type parameters (#404) (@baumgold)
• enable incremental reads of arrow-formatted files (#408) (@baumgold)
• Base.isdone for Stream (#428) (@baumgold)
• Run with 1 and 2 threads during tests (#431) (@quinnj)
• Add handling of len = -1 in uncompress (#436) (@DrChainsaw)
• Don't treat Vector{UInt8} as Arrow Binary type (#439) (@quinnj)
• Bump BitIntegers compat (#441) (@quinnj)
• Handle len of -1 in "compressed" buffers from other languages (#442) (@quinnj)
• Add Tables.partitions definition for Arrow.Table (#443) (@quinnj)
• Remove scopedenum for EnumX (#444) (@quinnj)
• Refactor compressors/decompressors for laziness + safety (#445) (@quinnj)
• Return SubArrays when possible for arrow list types (#446) (@quinnj)
• bump version of Arrow and ArrowTypes to prepare for new release (#447) (@baumgold)