Earley now uses OrderedSet for better output stability

[erezsh]

Addresses issue #1325, and others

[erezsh]

@MegaIng @chanicpanic This change makes Earley approx 10% slower. (on a very simple benchmark)

Do you think the stability it provides is worth it? (let's assume it gives us 100% output stability)

[MegaIng]

Maybe have an option that can change it, but default to the slower, stable version?

Also, I didn't check that this actually hits all locations, I assume you just searched for references to set.

[chanicpanic]

I like the idea of having an option to enable/disable it.

Defaulting to the stable version would be more user-friendly. The slowdown is unfortunate, but it is consistent with Lark positioning Earley as the parser for power and LALR as the parser for speed.

On the other hand, in addition to performance, it could be argued that defaulting to the unstable version has the benefit of exposing undesirable ambiguity and forcing grammar authors to remove/resolve it.

[codecov-commenter]

Codecov Report

Patch coverage: 92.98% and project coverage change: +0.02% 🎉

Comparison is base (c2e44ea) 89.23% compared to head (41c28d4) 89.26%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1327      +/-   ##
==========================================
+ Coverage   89.23%   89.26%   +0.02%     
==========================================
  Files          51       51              
  Lines        7591     7612      +21     
==========================================
+ Hits         6774     6795      +21     
  Misses        817      817              

Flag	Coverage Δ
unittests	89.26% <92.98%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
lark/load_grammar.py	92.60% <50.00%> (ø)
lark/parsers/earley_forest.py	77.94% <87.50%> (+0.19%)	⬆️
lark/utils.py	86.66% <93.75%> (+0.51%)	⬆️
lark/parsers/earley.py	82.65% <94.73%> (+0.57%)	⬆️
lark/lark.py	85.66% <100.00%> (+0.04%)	⬆️
lark/parser_frontends.py	95.90% <100.00%> (ø)
lark/parsers/xearley.py	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[erezsh]

@MegaIng @chanicpanic take 2