GTG does not use prepared datasets

[kocio-pl]

Launching GTG with -s should invoke two pre-made datasets, as mentioned in the README: bryce and screenshots. GTG creates datasets with requested name, but does not fill it with prepared data.

[picsel2]

How did you launch GTG? Did you use the launch.sh script?

[kocio-pl]

Yes, this is how I'm testing current version.

[picsel2]

It looks like current GTG is searching for a file named gtg_data.xml. The datasets in data/test-data/ looks like they are in an older data format.

This leads to GTG executing firstrun_tasks.generate unexpectedly.

[nekohayo]

Yeah, those datasets are in the old v1 (well, unversioned) format and haven't been fixed up and recommitted in clean v2 (0.5/0.6?) format, nor "converted to what 0.7 is" (if 0.7 actually is a different format? I'm not sure). I would actually like them to be present in all versions of the format, for (regression) testing purposes, if possible. @diegogangl does current git master / 0.7 use a different file format, or it remains strictly the same thing as the "v2" format?

[kocio-pl]

I almost finished converting script, but it looks like bryce dataset has multiple parents problems, which caused #855 warnings in the past and now trying to open this dataset fail because of them.

[kocio-pl]

The initial script is here:
https://github.com/kocio-pl/gtg/blob/old_datasets_conversion_script/scripts/convert_old_datasets.py

When run in the folder with old data files, it builds gtg_data.xml which is not too pretty, but both bryce and screenshots at least are opening with current app code.

Known problems:

• the tag counters are not always accurate (which might be at least in part a problem with app code, who knows?)
• some tasks include unconverted subtasks and tags in the content field (screenshots)
• task with multiple parents in bryce are just mechanically deparented, so no task has more than 1 parent now, but it would be good to check that no task has lost all the parents it had before

There might be even more problems with old datasets, but it's hard to say how much we want to care for them.