Cursively VS CsvBenchmarks

I was looking at the CSV parser Cursively recently, and I think it is a good simple example of a high performance C# parser and API design.

https://github.com/airbreather/Cursively

Agreed, and agreed. #21 and #22 seek to address this, but these have actually been very low priority for me: as your benchmarks show, if you primarily need a bunch of objects that must be UTF-16 strings, then are other libraries out there that will do the job just fine. The main reason to use Cursively for that would be if you have some use cases where you need the unusual qualities that Cursively offers, but other use cases where you can live with something more traditional, and you don't want to have two different CSV processing libraries.

The usage instructions are in the README on https://github.com/airbreather/Cursively. The most straightforward way to get started (for now) is:

The Sylvan.Data.Csv library is also the fastest CSV parser for .NET, and can load data about 3x faster than MTD based on my limited testing.

I don't want to disparage NPOI, because I'm sure it's great for its primary use-case, which is creating/editing spreadsheets. However, from a performance perspective it is one of the slowest libraries for reading Excel data in .NET. Since the OP seemed to be very concerned about performance and only needs to read data, this probably isn't the best choice.

I'm pretty sure NPOI doesn't support .xlsb. At least it didn't the last time I updated my benchmarks: https://github.com/MarkPflug/Benchmarks/blob/main/docs/ExcelBenchmarks.md

If you're trying to load an Excel worksheet into a DataTable, I maintain a library that can make this quite easy: Sylvan.Data.Excel. It is available as a nuget package, open source, MIT licensed, and the fastest Excel reader for .NET.

I maintain a library, Sylvan.Data.Excel that might work for you. I claim it is the fastest Excel library in the .NET ecosystem. It is open source and MIT licensed, so feel free to use it however you wish. It has no external dependencies other than .NET itself.

It provides a forward only data reader (DbDataReader) for all common excel formats (.xlsx, .xlsb, and .xls). It is the fastest and lowest allocating Excel reader in the .NET ecosystem.

This one is one that I personally maintain: Sylvan.Data.Excel It isn't fully featured, but it is the fastest that I'm aware of.

I maintain a couple libraries that should make this quite easy: Sylvan.Data.Csv, and Sylvan.Data.Excel. They are both open-source and MIT licensed. They are also both fastest-in-class at their particular tast.

It isn't exactly a full-featured library, but it is cross-platform and is the fastest and lowest allocating Excel reader that I'm aware of for .NET. That might not be a huge concern if you're just processing the occasional file.

Thanks for this -- I've pushed a PR that corrects a slight incorrectness issue that biased the results in Cursively's favor, uses the original byte array directly instead of needlessly wrapping it in a MemoryStream, and uses the same optimization that you used to make SylvanSimplePool win in its category. CursivelyCsv is now the winner on my machine in this category when doPooling is on.