MAINT: remove the Rust dependency

[misialq]

This PR removes the fastq_writer to get rid of the Rust dependency for our conda builds.

[misialq]

Hey @ebolyen, hi @lizgehret, is this what you had in mind or can we do it better? 😆

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (b418f80) 98.74% compared to head (ab2716f) 98.73%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #164      +/-   ##
==========================================
- Coverage   98.74%   98.73%   -0.01%     
==========================================
  Files          29       29              
  Lines        3031     3012      -19     
==========================================
- Hits         2993     2974      -19     
  Misses         38       38              

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

[lizgehret]

On first glance, this all seems reasonable to me @misialq!

EDIT: That being said, it might be worth adding a couple of unit tests just to make sure we're getting the same output as before.

[misialq]

Hey @ebolyen, thanks for your suggestions! I did try to change the code accordingly, but something strange is happening here... When I compare the file before and after compression (by "eye" and programmatically) they look identical. And yet QIIME keeps complaining that something is not quite right with the new one - here is the message I get when one of the tests fails:

Quality score length doesn't match sequence length for record beginning on line 13

I get a similar one when I just try to import that file into an artifact from the CLI. Is there anything obvious I'm missing here? I have not yet looked at what the validator does exactly...

[ebolyen]

Hi @misialq,

This code is correct, although there's a tiny strangeness to either the test-file or our validator, depending on perspective.

The original code did newline processing, and so would politely terminate the file with a final newline (you can run cat on the test file to see your shell bleed onto the last line).

Our validator is comparing line length, and includes the newline in its count. So:

ACGT\n  # this is 5 chars

85)9   # no trailing newline on the last line, so 4 chars

---

The super easy fix is to just add one more line to your test file. I don't think SRA will give you a non-newline terminated file, so I think just adding the newline to the test file is the way to go here.

The less easy fix is to update the validator to ignore newlines without adding too much more processing time (so maybe end-of-file only).

Alternatively we update this code to tack on a final newline if it's missing pre-gzip, but I don't think there's any nice way to do that.

[misialq]

Hey @ebolyen, I knew your keen eye would spot it! Thanks! That makes a lot of sense. I decided to go for the first and easiest solution, simply because (as you pointed out) the sequences fetched from SRA do have those newlines. Let me know whether that works for you and whether there are any other changes you'd like to see here. Otherwise, I'd be ready to merge. Thanks a lot!

[ebolyen]

Sorry this took me a minute to get back to! I've been swamped. This looks great, and we should be able to get it into our distros now :)