prepare as a python package

[davanstrien]

No description provided.

[andrewphilipsmith]

What format of python package do we which to use for publishing alto2txt?

• The README.md recommends installing in an Anaconda environment. As far as I can tell, there aren't any conda specific dependencies (I haven't checked any of the Spark-related code). If we recommend conda, then should we create a conda package and publish to conda-forge?
• If we publish to PyPI, should we remove the conda recommendation in the README?
• Is there a use case for publishing to both PyPI and conda-forge?

Apologies if I've missed something obvious here.

[mialondon]

Does it matter which version of ALTO files were created in? e.g. some parts of the BL are moving to 4.2, but there's a lot of material in v2.

[davanstrien]

Does it matter which version of ALTO files were created in? e.g. some parts of the BL are moving to 4.2, but there's a lot of material in v2.

You can see the current XSLTs here https://github.com/Living-with-machines/alto2txt/tree/master/extract_text/xslts. Depending on the changes in 4.2 it might be quite simple to adapt existing ones or it might be a lot of work. I won't have time to look into this soon. @GiorgiatolfoBL has a much better grasp on XML than me so might be able to give a sense of what changes (if any) are involved.

[andrewphilipsmith]

Does it matter which version of ALTO files were created in? e.g. some parts of the BL are moving to 4.2, but there's a lot of material in v2.

I don't think it matters for this issue. My question was intended to be about the options for publishing alto2txt itself as a software package. I'll reword it to make it clearer.

However, within the readme/other documentation for alto2txt, it should be explicit which versions of ALTO are and are not supported. Also, within issue #18, we should include tests that cover XML produced by different versions of ALTO.

[griff-rees]

The package is up on PyPI: https://pypi.org/project/alto2txt/ which may be enough to close this. If there are objections to its current release on ticket #48 that has implications here.