Repository navigation

Parallel sheet reads from one reader, positional sources and reader lifetime (v5.2.0) - #135

Merged

GabrielMarquezMatte merged 29 commits into

masterfrom

develop

Oct 6, 2026

Commits on Oct 5, 2026
- feat: enhance support for various Arrow types and improve handling of empty rows in readers
  GabrielMarquezMatte
  committed
  8379db3
Commits on Oct 6, 2026
- feat(core): add ByteSource, a positional view over memory, files and seekable streams
  
  GabrielMarquezMatte
  and
  claude
  committed
  74050b3
- feat(core): add ByteSourceStream, a buffered range stream over a ByteSource
  
  GabrielMarquezMatte
  and
  claude
  committed
  a3a3752
- refactor(core): rename ZipMemoryIndex to ZipIndex and index names into the central directory
  
  GabrielMarquezMatte
  and
  claude
  committed
  702e21a
- feat(core): let ZipIndex read its directory and entries through any ByteSource
  
  GabrielMarquezMatte
  and
  claude
  committed
  40b7a04
- fix(core): reject stored ZIP entries whose sizes disagree
  
  GabrielMarquezMatte
  and
  claude
  committed
  5f79408
- refactor(core): read XLSX from streams and files through ZipIndex
  
  GabrielMarquezMatte
  and
  claude
  committed
  9492ec3
- refactor(core): read XLSB from streams and files through ZipIndex
  
  GabrielMarquezMatte
  and
  claude
  committed
  676d146
- refactor(core): detect ZIP workbooks with ZipIndex and drop ZipArchive from the read path
  
  GabrielMarquezMatte
  and
  claude
  committed
  38b7a10
- refactor(core): load XLS sectors through ByteSource instead of seeking a shared stream
  
  GabrielMarquezMatte
  and
  claude
  committed
  fc38fdf
- fix(core): accept bytes after the ZIP end-of-central-directory record
  
  GabrielMarquezMatte
  and
  claude
  committed
  696134e
- test(core): cover ZIP64, async and failure paths of the positional ZIP index
  
  GabrielMarquezMatte
  and
  claude
  committed
  f03340e
- fix(core): restore stream position on failed detection and wrap only exact stream types
  
  GabrielMarquezMatte
  and
  claude
  committed
  fd30186
- test(core): make the inverting MemoryStream subclass not recurse
  
  GabrielMarquezMatte
  and
  claude
  committed
  d408b0f
- fix(core): count decompressed bytes atomically
  
  GabrielMarquezMatte
  and
  claude
  committed
  08933ce
- refactor(core): publish cached shared strings with compare-and-swap
  
  GabrielMarquezMatte
  and
  claude
  committed
  0e99b30
- feat(core): add OnceGate and ReaderLifetime for shared reader state
  
  GabrielMarquezMatte
  and
  claude
  committed
  e686e73
- feat(core): load XLSX shared strings once and keep the reader alive for its enumerators
  
  GabrielMarquezMatte
  and
  claude
  committed
  1e49a2b
- feat(core): keep an XLSB reader alive for its enumerators
  
  GabrielMarquezMatte
  and
  claude
  committed
  fecefc4
- feat(core): keep an XLS reader alive for its enumerators
  
  GabrielMarquezMatte
  and
  claude
  committed
  c8dbaa6
- fix(core): refuse use after dispose in the sources and reject negative offsets everywhere
  
  GabrielMarquezMatte
  and
  claude
  committed
  5f745f1
- fix(core): dispose the row enumerator when the ExcelDataReader constructor throws
  
  GabrielMarquezMatte
  and
  claude
  committed
  cee9ea5
- test(core): read sheets of one reader in parallel across formats and sources
  
  GabrielMarquezMatte
  and
  claude
  committed
  6256c76
- fix(core): stop a disposed row enumerator from reopening its sheet
  
  GabrielMarquezMatte
  and
  claude
  committed
  d157b8f
- test(core): cancel an XLSX shared-string load and load again
  
  GabrielMarquezMatte
  and
  claude
  committed
  ea28790
- feat(core): read CSV positionally and keep the reader alive for its enumerators
  
  GabrielMarquezMatte
  and
  claude
  committed
  853f168
- fix(core): offer no raw chunk source for a transcoded CSV file
  
  GabrielMarquezMatte
  and
  claude
  committed
  ea85fef
- perf(core): read CSV sources straight into the enumerator buffer
  
  GabrielMarquezMatte
  and
  claude
  committed
  c773e7d
- Merge branch 'master' into develop
  GabrielMarquezMatte
  authored
  1723f07

Navigation Menu

Repository navigation

Parallel sheet reads from one reader, positional sources and reader lifetime (v5.2.0) - #135

Parallel sheet reads from one reader, positional sources and reader lifetime (v5.2.0)#135
GabrielMarquezMatte merged 29 commits into
masterfrom
develop

Commits on Oct 5, 2026

feat: enhance support for various Arrow types and improve handling of empty rows in readers

Commits on Oct 6, 2026

feat(core): add ByteSource, a positional view over memory, files and seekable streams

feat(core): add ByteSourceStream, a buffered range stream over a ByteSource

refactor(core): rename ZipMemoryIndex to ZipIndex and index names into the central directory

feat(core): let ZipIndex read its directory and entries through any ByteSource

fix(core): reject stored ZIP entries whose sizes disagree

refactor(core): read XLSX from streams and files through ZipIndex

refactor(core): read XLSB from streams and files through ZipIndex

refactor(core): detect ZIP workbooks with ZipIndex and drop ZipArchive from the read path

refactor(core): load XLS sectors through ByteSource instead of seeking a shared stream

fix(core): accept bytes after the ZIP end-of-central-directory record

test(core): cover ZIP64, async and failure paths of the positional ZIP index

fix(core): restore stream position on failed detection and wrap only exact stream types

test(core): make the inverting MemoryStream subclass not recurse

fix(core): count decompressed bytes atomically

refactor(core): publish cached shared strings with compare-and-swap

feat(core): add OnceGate and ReaderLifetime for shared reader state

feat(core): load XLSX shared strings once and keep the reader alive for its enumerators

feat(core): keep an XLSB reader alive for its enumerators

feat(core): keep an XLS reader alive for its enumerators

fix(core): refuse use after dispose in the sources and reject negative offsets everywhere

fix(core): dispose the row enumerator when the ExcelDataReader constructor throws

test(core): read sheets of one reader in parallel across formats and sources

fix(core): stop a disposed row enumerator from reopening its sheet

test(core): cancel an XLSX shared-string load and load again

feat(core): read CSV positionally and keep the reader alive for its enumerators

fix(core): offer no raw chunk source for a transcoded CSV file

perf(core): read CSV sources straight into the enumerator buffer

Merge branch 'master' into develop