AttributeDict centralization about pypiper HOT 5 CLOSED

@vreuter what do you think of making AttributeDIct a separate project and putting it on pypi?

Actually, looks there someone else has already done something similar... https://pypi.org/project/attributedict/

And here's a more popular, older project that does the same thing: https://pypi.org/project/attrdict/

any chance to we can use those...?

from pypiper.

Yeah, this has been on my radar for awhile.
for now let's just leave them duplicated since it's working and I don't want to open this box quite yet.

long term I think the answer is that AttributeDict will reside in models, and both looper and pypiper will import it from models. So this will happen at the models/looper division.

from pypiper.

related to #68

from pypiper.

Definitely in favor of centralizing. I think the implementation used in this project may have slightly different default behavior regarding things like the "identity" return for missing attributes (since, e.g., requesting samtools and returning "samtools" is more likely to make sense in the pipeline context than in the project context.) That's just something that we should be aware/careful of when this is done.

Also fine with moving to an external implementation like you suggest. There are some custom behaviors like the identity thing just mentioned, and probably some others like path expansions that are automatic for us that we'd need to make accommodating changes for, or extend the functionality offered by an external project's data types.

from pypiper.

Closed in favor of #122

from pypiper.