Are there any plans to implement output (to bioinformatics formats)?

[J-Wall]

Great library, thanks for sharing this with the community. I'm just wondering if you have any plans to implement (write|sink)_(bed|vcf|bam|cram|fastq|fasta|gff3)?

I'm not sure why this looks like from an API standpoint. AFAIK you can't (and probably shouldn't) add these as methods to polars.LazyFrame, but maybe they could just be functions which accept a LazyFrame (for sink_ functions, or DataFrame for the write_ functions). I suppose these would need to validate that the schema is correct for the given output format.

As noted by @UriNeri in #209, currently file I/O is currently focused on inputs.

[mwiewior]

@J-Wall, thank you for your interest in polars-bio. The short answer: yes, refactoring the coordinate system for better consistency and streaming output formats are planned for 2026Q1!

[mwiewior]

hey @J-Wall and @wangzc66 I'm about to start working on this issue - could you pls let me know if you have priorities - which formats you would like to be implemented as first? I plan to start with FASTQ and VCF.

[J-Wall]

Hi @mwiewior, prioritise as you see fit! Funnily enough, in the meantime I implemented FASTA and FASTQ streaming output (and also slightly-hacky input) using purely polars LazyFrame expressions and the built in scan_csv and sink_csv. I have not yet benchmarked the scan_ functions against polars-bio's implementations, but one nice thing about them is they have no dependencies except polars, which makes it slightly easier to install on any version of python. Either way I'd be happy to contribute them to polars-bio if you are interested

[mwiewior]

hey @J-Wall, thanks! you may also take a look at our blog that summarizes some limitations of built-in scan_csv with handling compressed files.

[J-Wall]

Thanks @mwiewior. I can see you've already made some great progress on this. I'm really interested in how you go with bam/cram when you get to it!

[mwiewior]

#283 #276

[mwiewior]

Fixed.