Write unittests

[althonos]

Overview

Since the current build can take up to more than an hour, it would be preferable to have more fine-grained tests to use with mzml2isa. While we want to ensure compatibility with MetaboLights studies, it is useless to generate ISA files for the whole set of mzML files.

A new design of test follows that I'll work on implementing in the feat-tests branch.

Design:

using unittest as a testing framework:

File: test_mzml

Test behaviour of mzml2isa.mzml.mzMLmeta on an unique, embedded mzML file

File: test_imzml

Test behaviour of mzml2isa.mzml.imzMLmeta on an unique, embedded imzML file

File: test_mtbls

• if not in CI: test for all metabolights studies in ftp.ebi.ac.uk/pub/databases/metabolights/public/studies using pyfilesystem2 to provide an interface to the files without downloading them.
• if in CI: do the same thing but only for the study set in the Travis-CI environment (so builds would not attempt converting the whole MTBLS datasets)

File: test_api

Test functions exposed by the mzml2isa.parsing module

File: test_cli

Test command line interface of mzML2isa with several argument values

File: test_usermeta_loader

Test importing of piped json, json file and xlsx file containing metadata

New requirements:

• pyfilesystem2 (test only): not distributed on PyPI yet, i'd have liked to at least pin it but since they didn't do a single tag we'll have to set pip to install https://github.com/PyFilesystem/pyfilesystem2/archive/master.zip

[agbeltran]

@althonos unit tests should usually be run in a few seconds (irrespective if it is in CI or not). I was wondering why you think it is important to run the test over all the Mtbls datasets? this could be available as part of the functionality instead

[althonos]

Parsing on all metabolights datasets is mandatory as most mzML files are not following the standard format defined by the HUPO-PSI, so we end up having lots of slight differences that could make the program fail.

I implemented a long testing series to check that mzml2isa would AT LEAST not fail on MTBLS publicly available mzml.

What i will probably be doing instead is just checking 1 file per study as all mzml files in the same study should follow the same (maybe nonstandard) format.

[althonos]

Implemented in v1.0.0.