Test Sparv on Windows

[anne17]

Check if it's possible to install and run the pipeline on a Windows machine.

• [x] Test installing Sparv
• [x] Run small test corpus
• [ ] Run complex test corpora using all third-party software
• [ ] Document the things that probably won't work (due to dependencies to unix command line tools)
• [ ] Test how Sparv handles documents with Windows line breaks and the preserved_format export!

[anne17]

Testing Sparv on Windows 10 with Python 3.8

After installing pipx and re-booting my computer, installation of sparv-pipeline worked without problems. 🎉

1. Problems with running Sparv

1.1 [SOLVED] When trying to run sparv config I got an error message and the terminal stopped working. Martin figured out that the problem is in log_handler line 330 (log_file.unlink()). Sparv is trying to remove Snakemakes log file before Snakemake is done. Can be avoided by handling the PermissionError.

1.2 [DONE] The hunpos binaries end on ".exe" while Sparv is looking for binaries without the extension. This is solved by setting hunpos.binary: hunpos-tag.exe. Document this!

1.3 [DONE] I couldn't run hunpos because I was missing cygwin1.dll. Solution: Hunpos is shipped with the file. Just copy it into your sparv bin dir. Document this!

1.4 to be continued...

2. Other notes (less problematic)

2.1 The output from sparv annotations looses most of its markup on Windows, making it slightly harder to read.

2.2 Unicode symbols are not displayed which makes the progress bar look quite ugly.

2.3 Alive-progress seems to crop too much of the output in the progess bar.