0.10.0 grammer, elasticsearch transformer, setuptools extra

[markus1978]

Sorry for doing multiple things at the same time. But it is all kinda connected.

This PR contains:

• 0.10.0 grammer file
• elasticsearch transformer (only for 0.10.0)
• an 'extra' server (as in pip install optimade[server]) that omits all server/transformer related requirements

Unfortunately, I had to break compatibility with older grammars. The old grammars put AND and OR on the same precedence in contradiction to the spec. The necessary new grammar rules will break transformers on the old grammars.

The tests do not really run the queries on elasticsearch, they only parse the example queries and transform them. I don't have the time to get into the necessary travis ci/elasticsearch setup. Unfortunately, there is no easy mongomock kinda thing for elasticsearch.

[codecov]

Codecov Report

Merging #63 into master will increase coverage by 1.88%.
The diff coverage is 87.56%.

@@            Coverage Diff             @@
##           master      #63      +/-   ##
==========================================
+ Coverage   78.17%   80.06%   +1.88%     
==========================================
  Files          21       23       +2     
  Lines         724      908     +184     
==========================================
+ Hits          566      727     +161     
- Misses        158      181      +23

Impacted Files	Coverage Δ
...imade/filtertransformers/tests/test_transformer.py	100% <100%> (ø)	⬆️
...ade/filtertransformers/tests/test_elasticsearch.py	84.61% <84.61%> (ø)
optimade/filtertransformers/elasticsearch.py	87.97% <87.97%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5e5daf0...a6045a9. Read the comment docs.

[markus1978]

I missed a dependency on the first try.

[ml-evs]

Thanks for this @markus1978! Similar to my review of the Django transformer, I've never used ElasticSearch so you're much better equipped to assess your own work, but again this looks really useful. Summary of my comments:

1. Could you contribute the grammar to the v0.10 PR [WIP] Adding grammar for v0.9.8 #52 please?
2. Could we divide the requirements up further into each backend?
3. Do we really need to depend on ASE?

[markus1978]

@ml-evs let me respond to your 3 summary points:

for 3. I am just using the ase.data.chemical_symbols list. I could simply copy it, I guess it is a pretty stable piece of information.

for 2. Can do

for 1. This is the big one. I started with the newest grammar in #52 (0.9.8), but it was not working for me. I guess it is mixing various lark versions? Anyhow, I could not resolve the issues in it. But, the bigger problem is that Transformers are tied to the grammar rules. In a way, you can't really design them independently. I happy to commit my grammar to other places, but I don't really want to adapt the Transformer to another grammar. I think, this will also become an issue for the other transformers, especially when the grammar becomes more and more complex. For example, I tried to make the grammar compatible with the older one, but simply could not find a way to do it. Maybe this even gets to the point, where you have multiple grammars, each per Transformer?

[ml-evs]

Hi @markus1978, thanks for your comments and sorry for the delay getting back to you. I've just made a PR #64 for point 2 that splits requirements up per backend (please feel free to review...). Once that's merged I can rebase your PR and switch the requirements around a bit.

Regarding 3. you're clearly correct that we need to have backend/transformer-dependent grammars. I'm happy to accept this PR if you make it clear that the grammar you've written is intended for the elasticsearch transformer, then hopefully we can figure out a more general solution at a later date.

[markus1978]

@ml-evs , I changed the grammar file name to v.0.10.0.elastic.g. This way, the grammar is still available via version=(0, 10, 0). Let me know, if you had something different in mind.

[ml-evs]

Thanks for making those changes @markus1978! Happy to accept.