Parallel typed CSV reads, text-based schema inference and in-memory encryption (v5.1.0) by GabrielMarquezMatte · Pull Request #131 · GabrielMarquezMatte/ExcelReader

GabrielMarquezMatte · 2026-09-28T18:53:07Z

Summary

Parallel typed CSV parsing. xl_parse_typed_ex / xl_parse_arrow_ex take a degree of
parallelism; Python's parse_typed, to_arrow, to_record_batch and to_polars expose it as
parallelism (1 = one thread, the default; 0 = every core). ~3.6x faster on a 160 MB CSV with
16 threads. Other formats, and CSVs under 256 KiB, still read on one thread with the same result.
Schema inference from text. Excel.InferSchema(..., parseText: true), xl_infer_schema_ex
with XL_INFER_PARSE_TEXT, and Python's infer_schema(parse_text=True) type text cells (every
CSV field) as integers, decimals, booleans and ISO dates/date-times when the text has exactly that
shape. Leading-zero codes, scientific notation, integers past long, padded or comma-decimal
numbers and non-ISO dates stay strings.
In-memory encryption and frame writers. xl_encrypt_package_to_memory and Python's
encrypt_package_bytes() encrypt a package without writing its plaintext to disk.
write_arrow_to_bytes, write_pandas_to_bytes and write_polars_to_bytes complete the
in-memory writers.
Performance. FastDouble parses integer and fraction digits in separate loops; the parallel
table's columns are built concurrently; ColumnBuilder.AppendFrom is a thin per-cell dispatcher
(-2.5% CSV, -2% XLSB sequential in AOT). The static PGO profile is regenerated.

Fixes

Partitioned CSV reads keep a BOM when the dialect does not detect it (also affects the existing
parallel CSV aggregation).
A sequential typed CSV parse no longer reopens the file for overlapped reads it never uses
(regression from this branch's parallel work).

Compatibility

No breaking changes. XL_ABI_VERSION stays 5; every new export is additive.
.NET: InferSchema(reader, headerRow, sampleSize) keeps its binary signature; the new overload
adds parseText = false.
Not in this PR: Rust and C++ wrappers for the _ex exports and xl_encrypt_package_to_memory
(the Rust .def is updated so the import library links).

Test plan

dotnet test tests/ExcelReader.Tests — 2444 passed
pytest python/tests against a freshly published native library — 123 passed, 1 xfailed (pre-existing)
AOT timings for CSV/XLSX/XLSB before and after the PGO regeneration (no format regressed)

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…ch and to_polars

…handle's row cursor Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…not detect it

…overlapped handle on Windows A synchronous handle serializes every read on its file object, so 16 chunk readers sharing the FileStream's handle queued behind each other (cell scan 52 -> 25 ms on a 160 MB file). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…tion's columns The partitions were merged into the first one's builders on one thread (53 ms of a 160 MB parse) before BuildTable copied the result again. BuildTable now concatenates every partition's buffers into the native block once, rebasing string offsets and splicing validity bits. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…e tight loops The single loop checked for the dot and for leading zeros on every digit. Significant digits are now only recounted when the raw digit count exceeds the limit, and Pow10 is a table. Identical results to the old parser over 20M random inputs; ~3-4% on sequential CSV and XLSX typed parses through the AOT library.

…rewrite

…eText)

…on one thread xl_parse_typed and xl_parse_arrow now route through ParseTypedTable with a degree of parallelism of 1, but the chunk source was resolved before the partition check, so every file-backed CSV parse on Windows reopened the file for overlapped reads it never used. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

FastDouble accepts an exponent, so parseText typed a column of part codes such as 12E4 as Float64 and parse_typed then silently read it as 120000. Text with an e or E now stays a string, like leading-zero codes do. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…and its _ex twin xl_parse_typed/xl_parse_typed_ex and xl_parse_arrow/xl_parse_arrow_ex were copies apart from the degree of parallelism. Each pair now forwards to one private helper, and the arrow path takes its cause from the caller, so xl_parse_arrow_ex names itself in a live-session error like xl_parse_typed_ex. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

After the partitions are parsed, copying every column into its native block ran on one thread and took about a quarter of a parallel parse. Columns now build under Parallel.For when the parse itself ran in parallel; a sequential parse keeps the single-threaded loop. On a 521 MB CSV with 16 threads the parse drops from ~370 ms to ~330 ms read from a file and to ~300 ms read from memory. The columns block is zeroed first so a failed build frees whichever columns finished, and the allocation hooks move from ThreadStatic to AsyncLocal so fault-injection tests still reach the pool threads. Merging a byte-aligned validity bitmap also copies straight into the destination instead of through a temporary array. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

The double parser was inlined into AppendFrom, so every cell of every column type paid its eight-register prologue and stack zeroing just to be dispatched. Int64, Float64 and Bool now append through their own methods, and RecordValidity's no-null fast path is inlined. With the static PGO profile regenerated for the new methods, an AOT typed parse gets ~2.5% faster on CSV and ~2% on XLSB; XLSX is unchanged. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…t column build Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

xl_encrypt_package only takes file paths, so encrypting a workbook meant writing its plaintext package to disk first. The new export takes the package bytes and returns the encrypted workbook in an xl_buffer released with xl_free_buffer, so a package from xl_write_typed_to_memory never reaches disk. Both exports share the password length check. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

encrypt_package_bytes() encrypts a package's bytes through xl_encrypt_package_to_memory, so write_workbook_to_bytes() plus it produces an encrypted workbook without the plaintext touching disk. write_arrow_to_bytes(), write_pandas_to_bytes() and write_polars_to_bytes() are the in-memory twins of the frame writers, sharing their Arrow-to-table conversion with the file versions. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

…rceTests

codecov-commenter · 2026-09-28T19:05:25Z

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 94.75410% with 16 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.47%. Comparing base (d9477aa) to head (fc5b616).

Files with missing lines	Patch %	Lines
src/ExcelReader.Core/Reader/Csv/CsvReader.cs	57.89%	5 Missing and 3 partials ⚠️
src/ExcelReader.Native/Typed/TypedApi.Parallel.cs	90.47%	4 Missing and 2 partials ⚠️
src/ExcelReader.Native/Typed/TypedApi.cs	98.37%	2 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #131      +/-   ##
==========================================
+ Coverage   89.32%   89.47%   +0.15%     
==========================================
  Files         180      181       +1     
  Lines       13416    13611     +195     
  Branches     2462     2514      +52     
==========================================
+ Hits        11984    12179     +195     
  Misses        954      954              
  Partials      478      478

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

github-actions · 2026-09-28T19:10:53Z

Benchmark Results

Baseline: master d9477aaa, from the series published by benchmark-pages.yml. Both sides ran on a shared ubuntu-latest runner, so treat anything under ±10% as noise.

Moved by at least 10%

Benchmark	master	This PR	Delta
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsx_Sylvan(Rows: 65536)`	182.07 ms	291.28 ms	+60.0%
`ExcelReader.Benchmarks.ChunkedParseBenchmark.Batched_10k(Rows: 50000)`	8.87 ms	13.57 ms	+53.0%
`ExcelReader.Benchmarks.ChunkedParseBenchmark.WholeSheet(Rows: 50000)`	8.91 ms	13.22 ms	+48.5%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Encrypted_Stream`	44.36 ms	65.03 ms	+46.6%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Encrypted_Memory`	44.61 ms	64.95 ms	+45.6%
`ExcelReader.Benchmarks.ChunkedParseBenchmark.Batched_1k(Rows: 50000)`	9.05 ms	13.18 ms	+45.6%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Encrypted_Stream_VerifyIntegrity`	45.09 ms	65.13 ms	+44.4%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsx_ExcelReader_Prefetch(Rows: 65536)`	37.44 ms	53.75 ms	+43.6%
`ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark.Stored_EndToEnd(Rows: 65536)`	21.16 ms	30.16 ms	+42.5%
`ExcelReader.Benchmarks.NativeRowReadBenchmark.ReadAllDecoded`	59.12 ms	83.99 ms	+42.1%
`ExcelReader.Benchmarks.NativeTypedParseBenchmark.ParseTyped`	37.47 ms	52.81 ms	+40.9%
`ExcelReader.Benchmarks.NativeRowReadBenchmark.NextRowBlob`	39.02 ms	54.89 ms	+40.7%
`ExcelReader.Benchmarks.NativeRowReadBenchmark.NextRowDecoded`	43.58 ms	60.12 ms	+37.9%
`ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark.EndToEnd_Prefetch(Rows: 65536)`	37.14 ms	50.75 ms	+36.7%
`ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark.IndexParse_Utf8Parser(Rows: 65536)`	2.34 ms	3.14 ms	+34.5%
`ExcelReader.Benchmarks.NativeTypedParseBenchmark.ParseArrow`	39.61 ms	53.27 ms	+34.5%
`ExcelReader.Benchmarks.ReadBenchmark.OfficeIMO(Rows: 50000)`	165.34 ms	218.72 ms	+32.3%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsx_ExcelReader(Rows: 65536)`	47.95 ms	62.13 ms	+29.6%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Large_Encrypted_OpenOnly`	44.62 ms	57.09 ms	+28.0%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Plain_Stream`	59.68 µs	76.01 µs	+27.4%
`ExcelReader.Benchmarks.NativeRowReadBenchmark.ReadAllBlob`	45.71 ms	57.93 ms	+26.7%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsb_ExcelReader(Rows: 65536)`	42.91 ms	54.24 ms	+26.4%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsx_ExcelReader_Materialized(Rows: 65536)`	75.03 ms	94.64 ms	+26.1%
`ExcelReader.Benchmarks.ParseBenchmark.OfficeIMO(Rows: 50000)`	158.05 ms	198.35 ms	+25.5%
`ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark.EndToEnd(Rows: 65536)`	48.91 ms	61.33 ms	+25.4%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsb_ExcelReader_Materialized_Interned(Rows: 65536)`	65.74 ms	80.37 ms	+22.3%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsb_ExcelReader_Memory(Rows: 65536)`	42.66 ms	51.92 ms	+21.7%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsb_Sylvan(Rows: 65536)`	77.03 ms	93.71 ms	+21.6%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsx_ExcelReader_Materialized_Interned(Rows: 65536)`	74.94 ms	89.74 ms	+19.7%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsb_ExcelReader_Materialized(Rows: 65536)`	67.63 ms	80.23 ms	+18.6%
`ExcelReader.Benchmarks.ArrowConversionBenchmark.CsvTypedWithInference(Rows: 100000)`	13.96 ms	16.41 ms	+17.5%
`ExcelReader.Benchmarks.StringHeavyReadBenchmark.Xlsb_ExcelReader_Prefetch(Rows: 65536)`	34.93 ms	40.97 ms	+17.3%
`ExcelReader.Benchmarks.ArrowConversionBenchmark.CsvAllString(Rows: 100000)`	23.31 ms	27.16 ms	+16.5%
`ExcelReader.Benchmarks.CsvParallelVsSepBenchmark.Sep_Sequential(Corpus: NarrowInt)`	585.37 ms	680.15 ms	+16.2%
`ExcelReader.Benchmarks.CsvReadBenchmark.Sep(Rows: 50000)`	11.84 ms	13.72 ms	+15.9%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Large_Encrypted_Stream`	106.43 ms	123.12 ms	+15.7%
`ExcelReader.Benchmarks.ArrowConversionBenchmark.XlsbTyped(Rows: 100000)`	18.75 ms	21.61 ms	+15.3%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Encrypted_OpenOnly`	49.90 ms	57.25 ms	+14.7%
`ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark.SharedStringPrologue(Rows: 65536)`	17.72 ms	19.95 ms	+12.6%
`ExcelReader.Benchmarks.ReadBenchmark.OfficeIMOXlsb(Rows: 50000)`	15.76 ms	17.73 ms	+12.5%
`ExcelReader.Benchmarks.DataReaderBenchmark.DataReader_GetValue(Rows: 50000)`	11.59 ms	12.96 ms	+11.8%
`ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark.IndexParse_DigitLoop(Rows: 65536)`	1.73 ms	1.93 ms	+11.6%
`ExcelReader.Benchmarks.CsvParallelVsSepBenchmark.Sep_Parallel(Corpus: NarrowInt)`	455.35 ms	503.69 ms	+10.6%
`ExcelReader.Benchmarks.EncryptedWorkbookBenchmark.Large_Plain_Stream`	53.97 ms	59.54 ms	+10.3%
`ExcelReader.Benchmarks.ParseBenchmark.RefStructModelSync(Rows: 50000)`	15.79 ms	14.16 ms	-10.4%
`ExcelReader.Benchmarks.CsvParallelParseBenchmark.ConversionHeavy(Dop: 2)`	1.16 s	1.03 s	-11.1%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsb_ExcelReader_Prefetch`	19.88 ms	17.46 ms	-12.2%
`ExcelReader.Benchmarks.CsvReadBenchmark.ExcelReader(Rows: 50000)`	6.81 ms	5.98 ms	-12.2%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsb_ExcelReader_Memory_Prefetch`	20.08 ms	17.57 ms	-12.5%
`ExcelReader.Benchmarks.ParseBenchmark.ExcelParserSync(Rows: 50000)`	18.37 ms	15.95 ms	-13.2%
`ExcelReader.Benchmarks.CsvParseBenchmark.Sep(Rows: 50000)`	14.81 ms	12.72 ms	-14.1%
`ExcelReader.Benchmarks.ParseBenchmark.ExcelParserAsync(Rows: 50000)`	19.72 ms	16.72 ms	-15.2%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Csv_Sylvan`	6.40 ms	5.40 ms	-15.6%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsb_ExcelReader_Materialized`	37.89 ms	31.83 ms	-16.0%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Csv_ExcelReader_Materialized`	29.30 ms	24.31 ms	-17.0%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsb_ExcelReader`	35.99 ms	29.45 ms	-18.2%
`ExcelReader.Benchmarks.WritePathBenchmark.NativeStrings_Csv`	7.16 ms	5.85 ms	-18.3%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xls_ExcelReader_Materialized`	17.70 ms	14.44 ms	-18.4%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsb_ExcelReader_Memory`	35.70 ms	29.08 ms	-18.5%
`ExcelReader.Benchmarks.RealDataTypedParseBenchmark.Xlsb_ExcelParser`	47.16 ms	38.33 ms	-18.7%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsb_Sylvan`	43.94 ms	35.64 ms	-18.9%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xls_Sylvan`	26.95 ms	21.76 ms	-19.3%
`ExcelReader.Benchmarks.WriteBenchmark.ExcelReaderWriterPrefetch(Rows: 50000)`	12.82 ms	10.02 ms	-21.8%
`ExcelReader.Benchmarks.WriteBenchmark.OfficeIMO(Rows: 50000)`	24.52 ms	19.16 ms	-21.9%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xls_ExcelReader`	15.47 ms	12.05 ms	-22.1%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Csv_ExcelReader`	8.39 ms	6.50 ms	-22.5%
`ExcelReader.Benchmarks.CsvWriteBenchmark.ExcelReaderWriter(Rows: 50000)`	6.53 ms	5.03 ms	-23.1%
`ExcelReader.Benchmarks.WriteBenchmark.ExcelReaderXlsbWriter(Rows: 50000)`	10.67 ms	8.13 ms	-23.8%
`ExcelReader.Benchmarks.RecordWriteBenchmark.Xlsb(Rows: 50000)`	10.80 ms	8.23 ms	-23.8%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xls_ExcelReader_Memory`	15.65 ms	11.92 ms	-23.9%
`ExcelReader.Benchmarks.WriteBenchmark.ExcelReaderXlsbWriterSharedStrings(Rows: 50000)`	10.61 ms	8.07 ms	-23.9%
`ExcelReader.Benchmarks.RecordWriteBenchmark.Csv(Rows: 50000)`	7.05 ms	5.34 ms	-24.3%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsx_Sylvan`	308.96 ms	232.66 ms	-24.7%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsm_Sylvan`	309.24 ms	232.81 ms	-24.7%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsx_ExcelReader_Materialized`	83.08 ms	62.51 ms	-24.8%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsm_ExcelReader_Memory`	80.00 ms	60.01 ms	-25.0%
`ExcelReader.Benchmarks.RecordWriteBenchmark.XlsxMapped(Rows: 50000)`	18.80 ms	14.06 ms	-25.2%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsm_ExcelReader`	80.25 ms	59.86 ms	-25.4%
`ExcelReader.Benchmarks.WriteBenchmark.SpreadCheetah(Rows: 50000)`	21.89 ms	16.31 ms	-25.5%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsx_ExcelReader`	81.64 ms	60.74 ms	-25.6%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Csv_ExcelReader_Memory`	8.41 ms	6.25 ms	-25.6%
`ExcelReader.Benchmarks.CsvWriteBenchmark.Sep(Rows: 50000)`	10.39 ms	7.71 ms	-25.8%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsm_ExcelReader_Materialized`	83.07 ms	61.67 ms	-25.8%
`ExcelReader.Benchmarks.WritePathBenchmark.ArrowStrings_Xlsx`	36.72 ms	27.17 ms	-26.0%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsx_ExcelReader_Memory`	80.98 ms	59.82 ms	-26.1%
`ExcelReader.Benchmarks.RecordWriteBenchmark.XlsbMapped(Rows: 50000)`	12.34 ms	9.08 ms	-26.4%
`ExcelReader.Benchmarks.RealDataTypedParseBenchmark.Xlsx_ExcelParser`	96.28 ms	70.79 ms	-26.5%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsm_ExcelReader_Memory_Prefetch`	54.76 ms	40.01 ms	-26.9%
`ExcelReader.Benchmarks.RecordWriteBenchmark.CsvMapped(Rows: 50000)`	8.22 ms	6.00 ms	-27.0%
`ExcelReader.Benchmarks.WriteBenchmark.ExcelReaderWriterSharedStrings(Rows: 50000)`	16.04 ms	11.68 ms	-27.2%
`ExcelReader.Benchmarks.CsvWriteBenchmark.SylvanWriter(Rows: 50000)`	10.61 ms	7.66 ms	-27.9%
`ExcelReader.Benchmarks.RecordWriteBenchmark.Xlsx(Rows: 50000)`	17.60 ms	12.69 ms	-27.9%
`ExcelReader.Benchmarks.WritePathBenchmark.StyledRows_Xlsx`	10.30 ms	7.42 ms	-27.9%
`ExcelReader.Benchmarks.WritePathBenchmark.NativeStrings_Xlsx`	28.96 ms	20.85 ms	-28.0%
`ExcelReader.Benchmarks.XlsWriteBenchmark.XlsxWriter(Rows: 50000)`	16.17 ms	11.56 ms	-28.5%
`ExcelReader.Benchmarks.WriteBenchmark.ExcelReaderWriter(Rows: 50000)`	16.73 ms	11.50 ms	-31.2%
`ExcelReader.Benchmarks.WriteBenchmark.ExcelReaderXlsbWriterPrefetch(Rows: 50000)`	9.54 ms	6.36 ms	-33.3%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsx_ExcelReader_Memory_Prefetch`	57.58 ms	37.56 ms	-34.8%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsx_ExcelReader_Prefetch`	55.85 ms	35.97 ms	-35.6%
`ExcelReader.Benchmarks.XlsWriteBenchmark.XlsWriter(Rows: 50000)`	8.13 ms	5.01 ms	-38.3%
`ExcelReader.Benchmarks.RealDataReadBenchmark.Xlsm_ExcelReader_Prefetch`	55.96 ms	33.73 ms	-39.7%
`ExcelReader.Benchmarks.RecordWriteBenchmark.Xls(Rows: 50000)`	7.93 ms	4.62 ms	-41.7%

57 within ±10%.

Full results per group

ExcelReader.Benchmarks.ArrowConversionBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
INTEL XEON PLATINUM 8573C 2.30GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
CsvAllString	100000	27.16 ms	2.884 ms	0.749 ms	1.00	0.04	468.7500	468.7500	468.7500	16.27 MB	1.00
CsvTyped	100000	13.67 ms	0.815 ms	0.212 ms	0.50	0.01	453.1250	453.1250	453.1250	8.13 MB	0.50
XlsbTyped	100000	21.61 ms	0.242 ms	0.037 ms	0.80	0.02	437.5000	437.5000	437.5000	8.15 MB	0.50
CsvTypedWithInference	100000	16.41 ms	0.550 ms	0.143 ms	0.60	0.02	593.7500	593.7500	593.7500	14.89 MB	0.92

ExcelReader.Benchmarks.ChunkedParseBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.45GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	Gen0	Gen1	Allocated	Alloc Ratio
WholeSheet	50000	13.22 ms	0.107 ms	0.028 ms	1.00	78.1250	62.5000	1.31 MB	1.00
Batched_10k	50000	13.57 ms	0.087 ms	0.013 ms	1.03	93.7500	46.8750	1.6 MB	1.23
Batched_1k	50000	13.18 ms	0.047 ms	0.007 ms	1.00	93.7500	-	1.52 MB	1.16

ExcelReader.Benchmarks.ColdStartBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.45GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-GOEFXF : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

InvocationCount=1  IterationCount=1  LaunchCount=16  
RunStrategy=ColdStart  UnrollFactor=1  WarmupCount=0

Method	Rows	Mean	Error	StdDev	Allocated
TypedParseFirstUse	200	41.90 ms	1.312 ms	1.289 ms	28.43 KB
RecordWriteFirstUse	200	24.75 ms	1.707 ms	1.677 ms	84.25 KB
FluentParseFirstUse	200	36.12 ms	1.107 ms	1.088 ms	30.95 KB
FluentParseWithAttributeFallbackFirstUse	200	45.46 ms	0.302 ms	0.296 ms	32.65 KB

ExcelReader.Benchmarks.CsvParallelParseBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Dop	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Allocated	Alloc Ratio
ConversionHeavy	1	1,390.5 ms	44.51 ms	11.56 ms	1.00	0.01	42000.0000	-	687055.97 KB	1.000
NarrowInt	1	748.0 ms	27.61 ms	7.17 ms	0.54	0.01	15000.0000	-	251161.02 KB	0.366
ConversionHeavyAggregate	1	1,096.4 ms	52.52 ms	8.13 ms	0.79	0.01	-	-	826.38 KB	0.001
NarrowIntAggregate	1	693.2 ms	19.23 ms	2.98 ms	0.50	0.00	-	-	813.38 KB	0.001

ConversionHeavy	2	1,028.5 ms	27.56 ms	7.16 ms	1.00	0.01	42000.0000	23000.0000	694999.88 KB	1.000
NarrowInt	2	667.2 ms	20.74 ms	5.39 ms	0.65	0.01	15000.0000	3000.0000	259144.1 KB	0.373
ConversionHeavyAggregate	2	713.3 ms	76.53 ms	11.84 ms	0.69	0.01	-	-	1248.46 KB	0.002
NarrowIntAggregate	2	470.7 ms	25.79 ms	6.70 ms	0.46	0.01	-	-	1233.95 KB	0.002

ConversionHeavy	4	741.9 ms	69.99 ms	18.18 ms	1.00	0.03	42000.0000	20000.0000	695080.21 KB	1.000
NarrowInt	4	467.1 ms	30.43 ms	7.90 ms	0.63	0.02	15000.0000	1000.0000	259255.23 KB	0.373
ConversionHeavyAggregate	4	478.9 ms	45.64 ms	7.06 ms	0.65	0.02	-	-	1253.59 KB	0.002
NarrowIntAggregate	4	316.8 ms	27.38 ms	7.11 ms	0.43	0.01	-	-	1239.38 KB	0.002

ConversionHeavy	8	764.1 ms	85.82 ms	22.29 ms	1.00	0.04	42000.0000	21000.0000	695235.51 KB	1.000
NarrowInt	8	474.1 ms	66.12 ms	17.17 ms	0.62	0.03	15000.0000	1000.0000	259391.35 KB	0.373
ConversionHeavyAggregate	8	462.2 ms	17.97 ms	4.67 ms	0.61	0.02	-	-	1260.92 KB	0.002
NarrowIntAggregate	8	335.3 ms	84.12 ms	21.85 ms	0.44	0.03	-	-	1252.27 KB	0.002

ConversionHeavy	16	719.7 ms	56.59 ms	8.76 ms	1.00	0.02	42000.0000	20000.0000	695306.02 KB	1.000
NarrowInt	16	464.4 ms	24.86 ms	3.85 ms	0.65	0.01	15000.0000	1000.0000	260066.42 KB	0.374
ConversionHeavyAggregate	16	473.5 ms	19.98 ms	5.19 ms	0.66	0.01	-	-	1309.09 KB	0.002
NarrowIntAggregate	16	330.6 ms	76.35 ms	19.83 ms	0.46	0.03	-	-	1303.76 KB	0.002

ExcelReader.Benchmarks.CsvParallelVsSepBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Corpus	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
ExcelReader_Sequential	NarrowInt	704.7 ms	4.43 ms	0.69 ms	1.00	0.00	15000.0000	-	-	244.15 MB	1.00
ExcelReader_Parallel	NarrowInt	468.5 ms	32.29 ms	8.39 ms	0.66	0.01	15000.0000	1000.0000	-	253.33 MB	1.04
Sep_Sequential	NarrowInt	680.2 ms	26.71 ms	6.94 ms	0.97	0.01	15000.0000	-	-	244.15 MB	1.00
Sep_Parallel	NarrowInt	503.7 ms	154.74 ms	23.95 ms	0.71	0.03	16000.0000	12000.0000	1000.0000	251.43 MB	1.03

ExcelReader_Sequential	ConversionHeavy	936.2 ms	7.71 ms	1.19 ms	1.00	0.00	29000.0000	-	-	467.31 MB	1.00
ExcelReader_Parallel	ConversionHeavy	503.9 ms	46.63 ms	12.11 ms	0.54	0.01	29000.0000	26000.0000	-	473.66 MB	1.01
Sep_Sequential	ConversionHeavy	1,059.2 ms	71.24 ms	18.50 ms	1.13	0.02	29000.0000	-	-	467.31 MB	1.00
Sep_Parallel	ConversionHeavy	854.5 ms	63.60 ms	16.52 ms	0.91	0.02	36000.0000	34000.0000	6000.0000	476.99 MB	1.02

ExcelReader.Benchmarks.CsvParseBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	Gen0	Gen1	Allocated	Alloc Ratio
ExcelParserSync	50000	7.450 ms	0.0995 ms	0.0154 ms	1.00	234.3750	-	3.86 MB	1.00
ExcelParserAsync	50000	7.571 ms	0.0511 ms	0.0079 ms	1.02	234.3750	-	3.86 MB	1.00
Sep	50000	12.716 ms	0.1046 ms	0.0272 ms	1.71	234.3750	-	3.87 MB	1.00
Sylvan	50000	18.526 ms	0.3140 ms	0.0815 ms	2.49	656.2500	62.5000	10.95 MB	2.84

ExcelReader.Benchmarks.CsvReadBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.62GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Allocated	Alloc Ratio
ExcelReader	50000	5.984 ms	0.0299 ms	0.0078 ms	1.00	0.00	448 B	1.00
ExcelReaderWide	50000	12.525 ms	0.9844 ms	0.2556 ms	2.09	0.04	376 B	0.84
ExcelReaderAsync	50000	4.089 ms	0.0063 ms	0.0010 ms	0.68	0.00	520 B	1.16
Sep	50000	13.723 ms	0.2075 ms	0.0321 ms	2.29	0.01	4024 B	8.98
Sylvan	50000	5.933 ms	0.0220 ms	0.0034 ms	0.99	0.00	38682 B	86.34

ExcelReader.Benchmarks.CsvWriteBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
ExcelReaderWriter	50000	5.027 ms	0.5279 ms	0.1371 ms	1.00	0.03	500.0000	500.0000	500.0000	4 MB	1.00
Sep	50000	7.710 ms	0.1443 ms	0.0223 ms	1.53	0.04	500.0000	500.0000	500.0000	4.01 MB	1.00
SylvanWriter	50000	7.655 ms	0.1489 ms	0.0230 ms	1.52	0.04	500.0000	500.0000	500.0000	4.04 MB	1.01

ExcelReader.Benchmarks.DataReaderBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
INTEL XEON PLATINUM 8573C 2.30GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
Baseline_RawRows	50000	9.419 ms	0.0426 ms	0.0111 ms	1.00	0.00	-	-	-	11.18 KB	1.00
DataReader_GetValue	50000	12.958 ms	0.1453 ms	0.0377 ms	1.38	0.00	62.5000	-	-	5138.82 KB	459.66
DataReader_TypedGetters	50000	12.778 ms	0.0512 ms	0.0079 ms	1.36	0.00	15.6250	-	-	1623.2 KB	145.19
DataReader_GetBytes	50000	9.641 ms	0.0911 ms	0.0141 ms	1.02	0.00	-	-	-	11.87 KB	1.06
DataTable_Load	50000	65.154 ms	5.8148 ms	1.5101 ms	6.92	0.15	1111.1111	1000.0000	888.8889	23401.53 KB	2,093.22

ExcelReader.Benchmarks.EncryptedWorkbookBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Plain_Stream	76.01 μs	2.449 μs	0.379 μs	1.00	0.01	1.0986	19.42 KB	1.00
Encrypted_Stream	65,027.15 μs	2,737.971 μs	711.042 μs	855.51	9.38	-	48.33 KB	2.49
Encrypted_Stream_VerifyIntegrity	65,125.32 μs	567.065 μs	147.265 μs	856.80	4.18	-	50.27 KB	2.59
Encrypted_Memory	64,945.66 μs	689.442 μs	179.046 μs	854.43	4.35	-	52.77 KB	2.72
Encrypted_OpenOnly	57,245.07 μs	593.717 μs	154.186 μs	753.12	3.81	-	42.8 KB	2.20
Large_Plain_Stream	59,543.67 μs	1,598.143 μs	247.314 μs	783.36	4.54	-	11.18 KB	0.58
Large_Encrypted_Stream	123,118.33 μs	16,398.645 μs	4,258.676 μs	1,619.76	51.91	-	36.94 KB	1.90
Large_Encrypted_OpenOnly	57,086.35 μs	659.206 μs	102.013 μs	751.04	3.55	-	36.37 KB	1.87

ExcelReader.Benchmarks.NativeRowReadBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.45GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
NextRowBlob	54.89 ms	0.187 ms	0.029 ms	1.00	0.00	300.0000	300.0000	300.0000	3.71 MB	1.00
NextRowDecoded	60.12 ms	0.118 ms	0.018 ms	1.10	0.00	222.2222	222.2222	222.2222	3.71 MB	1.00
ReadAllBlob	57.93 ms	1.240 ms	0.192 ms	1.06	0.00	1000.0000	888.8889	-	21.06 MB	5.68
ReadAllDecoded	83.99 ms	5.227 ms	1.358 ms	1.53	0.02	500.0000	500.0000	500.0000	5.71 MB	1.54

ExcelReader.Benchmarks.NativeTypedParseBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.45GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Mean	Error	StdDev	Ratio	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
ParseTyped	52.81 ms	0.698 ms	0.181 ms	1.00	700.0000	500.0000	300.0000	11.61 MB	1.00
ParseArrow	53.27 ms	0.320 ms	0.050 ms	1.01	700.0000	500.0000	300.0000	11.62 MB	1.00

ExcelReader.Benchmarks.ParseBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
ExcelParserSync	50000	15.95 ms	0.078 ms	0.020 ms	1.00	0.00	218.7500	3966.8 KB	1.000
ExcelParserSyncSharedStrings	50000	15.45 ms	0.064 ms	0.010 ms	0.97	0.00	140.6250	2357.3 KB	0.594
ExcelParserStructSync	50000	15.40 ms	0.185 ms	0.048 ms	0.97	0.00	93.7500	1623.05 KB	0.409
RefStructModelSync	50000	14.16 ms	0.040 ms	0.006 ms	0.89	0.00	-	11.72 KB	0.003
ExcelParserAsync	50000	16.72 ms	0.062 ms	0.010 ms	1.05	0.00	218.7500	3968.95 KB	1.001
ExcelParserXlsbSync	50000	12.34 ms	0.472 ms	0.122 ms	0.77	0.01	234.3750	3968.9 KB	1.001
ExcelParserXlsbAsync	50000	12.36 ms	0.039 ms	0.010 ms	0.78	0.00	234.3750	3971.47 KB	1.001
OfficeIMO	50000	198.35 ms	136.593 ms	35.473 ms	12.44	2.03	-	5858.15 KB	1.477
Sylvan	50000	87.21 ms	15.328 ms	3.981 ms	5.47	0.23	-	10726.71 KB	2.704
SylvanAsync	50000	85.40 ms	5.602 ms	1.455 ms	5.36	0.08	-	10727.74 KB	2.704

ExcelReader.Benchmarks.ReadBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.62GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
ExcelReader	50000	11.427 ms	0.1088 ms	0.0168 ms	1.00	0.00	-	11.18 KB	1.00
ExcelReaderAsync	50000	12.409 ms	0.0323 ms	0.0050 ms	1.09	0.00	-	13.31 KB	1.19
ExcelReaderXlsb	50000	7.348 ms	0.2012 ms	0.0311 ms	0.64	0.00	-	13.28 KB	1.19
ExcelReaderXlsbAsync	50000	7.981 ms	0.0865 ms	0.0225 ms	0.70	0.00	-	15.84 KB	1.42
ExcelReaderMaterialized	50000	13.292 ms	0.0518 ms	0.0135 ms	1.16	0.00	93.7500	1622.51 KB	145.13
Sylvan	50000	55.772 ms	0.4923 ms	0.0762 ms	4.88	0.01	-	1939.18 KB	173.46
OfficeIMO	50000	218.723 ms	232.4554 ms	60.3679 ms	19.14	4.85	-	7003.68 KB	626.46
OfficeIMOXlsb	50000	17.729 ms	0.0734 ms	0.0191 ms	1.55	0.00	406.2500	6858.72 KB	613.50

ExcelReader.Benchmarks.RealDataReadBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Allocated	Alloc Ratio
Xlsx_ExcelReader	60.739 ms	4.5183 ms	1.1734 ms	1.00	0.02	-	18291 B	1.00
Xlsx_Sylvan	232.665 ms	1.7388 ms	0.4516 ms	3.83	0.07	-	659448 B	36.05
Xlsx_ExcelReader_Materialized	62.507 ms	0.8058 ms	0.1247 ms	1.03	0.02	-	27990 B	1.53
Xlsx_ExcelReader_Prefetch	35.970 ms	12.6361 ms	3.2816 ms	0.59	0.05	-	29596 B	1.62
Xlsx_ExcelReader_Memory	59.820 ms	0.2377 ms	0.0617 ms	0.99	0.02	-	7179 B	0.39
Xlsx_ExcelReader_Memory_Prefetch	37.559 ms	16.2277 ms	4.2143 ms	0.62	0.06	-	14867 B	0.81
Xlsm_ExcelReader	59.860 ms	0.3360 ms	0.0873 ms	0.99	0.02	-	19664 B	1.08
Xlsm_Sylvan	232.815 ms	9.6141 ms	2.4968 ms	3.83	0.08	-	659528 B	36.06
Xlsm_ExcelReader_Materialized	61.667 ms	0.3837 ms	0.0594 ms	1.02	0.02	-	27990 B	1.53
Xlsm_ExcelReader_Prefetch	33.728 ms	4.0393 ms	1.0490 ms	0.56	0.02	-	31377 B	1.72
Xlsm_ExcelReader_Memory	60.008 ms	0.5323 ms	0.1382 ms	0.99	0.02	-	7179 B	0.39
Xlsm_ExcelReader_Memory_Prefetch	40.012 ms	23.4334 ms	6.0856 ms	0.66	0.09	-	21400 B	1.17
Xlsb_ExcelReader	29.453 ms	0.4849 ms	0.0750 ms	0.49	0.01	-	19256 B	1.05
Xlsb_Sylvan	35.637 ms	1.3697 ms	0.3557 ms	0.59	0.01	-	346670 B	18.95
Xlsb_ExcelReader_Materialized	31.829 ms	0.1802 ms	0.0279 ms	0.52	0.01	-	28952 B	1.58
Xlsb_ExcelReader_Prefetch	17.463 ms	1.0287 ms	0.1592 ms	0.29	0.01	-	36098 B	1.97
Xlsb_ExcelReader_Memory	29.084 ms	0.1168 ms	0.0181 ms	0.48	0.01	-	8896 B	0.49
Xlsb_ExcelReader_Memory_Prefetch	17.574 ms	0.9866 ms	0.2562 ms	0.29	0.01	-	24628 B	1.35
Xls_ExcelReader	12.054 ms	0.2651 ms	0.0410 ms	0.20	0.00	-	11576 B	0.63
Xls_Sylvan	21.758 ms	0.3603 ms	0.0936 ms	0.36	0.01	-	190366 B	10.41
Xls_ExcelReader_Materialized	14.441 ms	0.7039 ms	0.1828 ms	0.24	0.00	-	21272 B	1.16
Xls_ExcelReader_Memory	11.918 ms	0.0048 ms	0.0007 ms	0.20	0.00	-	11576 B	0.63
Csv_ExcelReader	6.500 ms	0.1310 ms	0.0340 ms	0.11	0.00	-	376 B	0.02
Csv_Sylvan	5.402 ms	0.1680 ms	0.0436 ms	0.09	0.00	-	41226 B	2.25
Csv_ExcelReader_Materialized	24.313 ms	0.5176 ms	0.1344 ms	0.40	0.01	2218.7500	37442984 B	2,047.07
Csv_ExcelReader_Memory	6.254 ms	0.0131 ms	0.0020 ms	0.10	0.00	-	312 B	0.02

ExcelReader.Benchmarks.RealDataTypedParseBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Mean	Error	StdDev	Ratio	Gen0	Allocated	Alloc Ratio
Xlsx_ExcelParser	70.79 ms	1.872 ms	0.486 ms	1.00	428.5714	8.03 MB	1.00
Xlsb_ExcelParser	38.33 ms	0.691 ms	0.107 ms	0.54	500.0000	8.03 MB	1.00

ExcelReader.Benchmarks.RecordWriteBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
Xlsx	50000	12.688 ms	0.7088 ms	0.1841 ms	1.00	0.02	500.0000	500.0000	500.0000	4.02 MB	1.00
Xlsb	50000	8.234 ms	0.1197 ms	0.0311 ms	0.65	0.01	500.0000	500.0000	500.0000	4.02 MB	1.00
Xls	50000	4.619 ms	0.1693 ms	0.0440 ms	0.36	0.01	273.4375	273.4375	273.4375	4.03 MB	1.00
Csv	50000	5.337 ms	0.4115 ms	0.1069 ms	0.42	0.01	500.0000	500.0000	500.0000	4 MB	1.00
XlsxMapped	50000	14.060 ms	0.6694 ms	0.1738 ms	1.11	0.02	500.0000	500.0000	500.0000	4.02 MB	1.00
XlsbMapped	50000	9.082 ms	0.7437 ms	0.1931 ms	0.72	0.02	500.0000	500.0000	500.0000	4.02 MB	1.00
CsvMapped	50000	6.001 ms	0.1018 ms	0.0157 ms	0.47	0.01	500.0000	500.0000	500.0000	4 MB	1.00

ExcelReader.Benchmarks.StringHeavyReadBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.61GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
Xlsx_ExcelReader	65536	62.13 ms	1.544 ms	0.401 ms	1.00	0.01	-	-	-	2.19 MB	1.00
Xlsx_ExcelReader_Prefetch	65536	53.75 ms	25.857 ms	4.001 ms	0.87	0.06	-	-	-	2.25 MB	1.03
Xlsx_Sylvan	65536	291.28 ms	28.294 ms	7.348 ms	4.69	0.11	500.0000	-	-	17.41 MB	7.96
Xlsx_ExcelReader_Materialized	65536	94.64 ms	9.398 ms	2.441 ms	1.52	0.04	833.3333	666.6667	166.6667	15.26 MB	6.97
Xlsx_ExcelReader_Materialized_Interned	65536	89.74 ms	5.541 ms	0.857 ms	1.44	0.01	833.3333	666.6667	166.6667	15.26 MB	6.98
Xlsb_ExcelReader	65536	54.24 ms	0.733 ms	0.190 ms	0.87	0.01	-	-	-	2.19 MB	1.00
Xlsb_ExcelReader_Prefetch	65536	40.97 ms	2.069 ms	0.320 ms	0.66	0.01	-	-	-	2.3 MB	1.05
Xlsb_Sylvan	65536	93.71 ms	15.913 ms	2.463 ms	1.51	0.04	1166.6667	1000.0000	333.3333	17.38 MB	7.94
Xlsb_ExcelReader_Materialized	65536	80.23 ms	6.286 ms	1.633 ms	1.29	0.03	833.3333	666.6667	166.6667	15.26 MB	6.97
Xlsb_ExcelReader_Materialized_Interned	65536	80.37 ms	1.922 ms	0.297 ms	1.29	0.01	1000.0000	857.1429	285.7143	15.27 MB	6.98
Xlsb_ExcelReader_Memory	65536	51.92 ms	0.611 ms	0.095 ms	0.84	0.01	-	-	-	6.47 MB	2.95

ExcelReader.Benchmarks.WriteBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
ExcelReaderWriter	50000	11.505 ms	0.0585 ms	0.0090 ms	1.00	0.00	500.0000	500.0000	500.0000	4.02 MB	1.00
ExcelReaderWriterSharedStrings	50000	11.676 ms	0.0602 ms	0.0156 ms	1.01	0.00	500.0000	500.0000	500.0000	4.06 MB	1.01
ExcelReaderWriterPrefetch	50000	10.018 ms	1.0251 ms	0.2662 ms	0.87	0.02	484.3750	484.3750	484.3750	4.03 MB	1.00
ExcelReaderXlsbWriter	50000	8.132 ms	0.4537 ms	0.1178 ms	0.71	0.01	500.0000	500.0000	500.0000	4.02 MB	1.00
ExcelReaderXlsbWriterSharedStrings	50000	8.070 ms	1.2153 ms	0.3156 ms	0.70	0.03	500.0000	500.0000	500.0000	4.06 MB	1.01
ExcelReaderXlsbWriterPrefetch	50000	6.360 ms	0.7341 ms	0.1906 ms	0.55	0.02	492.1875	492.1875	492.1875	4.03 MB	1.00
OfficeIMO	50000	19.157 ms	1.2157 ms	0.3157 ms	1.67	0.03	312.5000	312.5000	312.5000	4.02 MB	1.00
SpreadCheetah	50000	16.307 ms	0.1793 ms	0.0466 ms	1.42	0.00	1437.5000	718.7500	718.7500	15.84 MB	3.94

ExcelReader.Benchmarks.WritePathBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Mean	Error	StdDev	Allocated
NativeStrings_Xlsx	20.849 ms	0.0884 ms	0.0137 ms	18448 B
NativeStrings_Csv	5.847 ms	0.0297 ms	0.0077 ms	416 B
ArrowStrings_Xlsx	27.173 ms	0.0885 ms	0.0230 ms	18496 B
StyledRows_Xlsx	7.424 ms	0.0434 ms	0.0067 ms	18632 B

ExcelReader.Benchmarks.XlsReadBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.62GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	Gen0	Gen1	Allocated	Alloc Ratio
ExcelReader	50000	4.964 ms	0.0109 ms	0.0028 ms	1.00	-	-	2.45 KB	1.00
ExcelReaderAsync	50000	5.181 ms	0.0330 ms	0.0086 ms	1.04	-	-	2.52 KB	1.03
Sylvan	50000	8.609 ms	0.2602 ms	0.0676 ms	1.73	93.7500	15.6250	1717.73 KB	700.22
OfficeIMO	50000	10.399 ms	0.1534 ms	0.0237 ms	2.09	640.6250	-	10499.25 KB	4,279.95

ExcelReader.Benchmarks.XlsWriteBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 9V74 2.60GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v4

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	RatioSD	Gen0	Gen1	Gen2	Allocated	Alloc Ratio
XlsWriter	50000	5.013 ms	0.0855 ms	0.0222 ms	1.00	0.01	492.1875	492.1875	492.1875	16.03 MB	1.00
XlsxWriter	50000	11.562 ms	0.3936 ms	0.1022 ms	2.31	0.02	500.0000	500.0000	500.0000	4.02 MB	0.25

ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark


BenchmarkDotNet v0.15.8, Linux Ubuntu 24.04.5 LTS (Noble Numbat)
AMD EPYC 7763 2.61GHz, 1 CPU, 4 logical and 2 physical cores
.NET SDK 10.0.401
  [Host]     : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3
  Job-MEHJPP : .NET 10.0.12 (10.0.12, 10.0.1226.42308), X64 RyuJIT x86-64-v3

IterationCount=5  WarmupCount=1

Method	Rows	Mean	Error	StdDev	Ratio	Allocated	Alloc Ratio
EndToEnd	65536	61.331 ms	0.3712 ms	0.0574 ms	1.00	2294168 B	1.00
EndToEnd_Prefetch	65536	50.753 ms	6.6185 ms	1.0242 ms	0.83	2391668 B	1.04
SharedStringPrologue	65536	19.949 ms	0.2593 ms	0.0674 ms	0.33	773304 B	0.34
Stored_EndToEnd	65536	30.158 ms	0.5432 ms	0.1411 ms	0.49	2293088 B	1.00
Stored_SharedStringPrologue	65536	4.872 ms	0.1677 ms	0.0435 ms	0.08	772224 B	0.34
IndexParse_Utf8Parser	65536	3.145 ms	0.0147 ms	0.0038 ms	0.05	-	0.00
IndexParse_DigitLoop	65536	1.925 ms	0.0322 ms	0.0084 ms	0.03	-	0.00

GabrielMarquezMatte and others added 29 commits September 25, 2026 23:45

feat(native): add ChunkedBuffer.AppendFrom

0ee80d7

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

feat(core): expose a CsvReader's partition source to the native layer

55efcbd

feat(native): parse CSV handles into typed tables in parallel

9f6c4cc

feat(native): export xl_parse_typed_ex and xl_parse_arrow_ex

14c9462

feat(python): add parallelism to parse_typed, to_arrow, to_record_bat…

f24b93c

…ch and to_polars

docs(python): document parallel CSV typed reads with measurements

59a4e81

fix(native): resolve the parallel parse header without moving a file …

ec3a225

…handle's row cursor Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

fix(core): keep a BOM in partitioned CSV reads when the dialect does …

216c56a

…not detect it

docs(python): re-measure parallel CSV typed reads

49e7a8c

feat(benchmarks): add CSV comparison for excelreader, pandas, and polars

8c915de

perf(native): regenerate the static PGO profile after the FastDouble …

85b1328

…rewrite

feat(core): infer column types from text cells behind a parseText option

b7e059b

feat(core): add Excel.InferSchema(reader, headerRow, sampleSize, pars…

cd95d3f

…eText)

feat(native): export xl_infer_schema_ex with XL_INFER_PARSE_TEXT

a1d68af

feat(python): add parse_text to infer_schema

de3eb45

fix(core): keep an integer past long as text when inferring from text

030de7c

Merge branch 'master' into develop

f5289e1

docs(python): re-measure parallel CSV typed reads after the concurren…

f81fafe

…t column build Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

fix(tests): clarify assertion for handle closure in CsvReaderChunkSou…

fc5b616

…rceTests

GabrielMarquezMatte merged commit 72c5781 into master Sep 28, 2026
58 checks passed

Conversation

GabrielMarquezMatte commented Sep 28, 2026

Summary

Fixes

Compatibility

Test plan

Uh oh!

codecov-commenter commented Sep 28, 2026

Codecov Report

Uh oh!

github-actions Bot commented Sep 28, 2026

Benchmark Results

Moved by at least 10%

ExcelReader.Benchmarks.ArrowConversionBenchmark

ExcelReader.Benchmarks.ChunkedParseBenchmark

ExcelReader.Benchmarks.ColdStartBenchmark

ExcelReader.Benchmarks.CsvParallelParseBenchmark

ExcelReader.Benchmarks.CsvParallelVsSepBenchmark

ExcelReader.Benchmarks.CsvParseBenchmark

ExcelReader.Benchmarks.CsvReadBenchmark

ExcelReader.Benchmarks.CsvWriteBenchmark

ExcelReader.Benchmarks.DataReaderBenchmark

ExcelReader.Benchmarks.EncryptedWorkbookBenchmark

ExcelReader.Benchmarks.NativeRowReadBenchmark

ExcelReader.Benchmarks.NativeTypedParseBenchmark

ExcelReader.Benchmarks.ParseBenchmark

ExcelReader.Benchmarks.ReadBenchmark

ExcelReader.Benchmarks.RealDataReadBenchmark

ExcelReader.Benchmarks.RealDataTypedParseBenchmark

ExcelReader.Benchmarks.RecordWriteBenchmark

ExcelReader.Benchmarks.StringHeavyReadBenchmark

ExcelReader.Benchmarks.WriteBenchmark

ExcelReader.Benchmarks.WritePathBenchmark

ExcelReader.Benchmarks.XlsReadBenchmark

ExcelReader.Benchmarks.XlsWriteBenchmark

ExcelReader.Benchmarks.XlsxSharedStringHotPathBenchmark

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants