Use Match-Replace strategy to alias operators

AlgebraicJulia / DiagrammaticEquations.jl

MIT License

12 stars 1 forks source link

Use Match-Replace strategy to alias operators #37

Closed lukem12345 closed 4 months ago

lukem12345 commented 5 months ago

Close #35

As a compromise between:

considerations of computational efficiency
warts relating to the types of "matched" variables destroying "context" , a "match & replace" strategy is adopted - inspired by AlgebraicRewriting - but using tight loops tailored to our workflow.

codecov[bot] commented 5 months ago

Codecov Report

Attention: Patch coverage is 79.41176% with 21 lines in your changes missing coverage. Please review.

Project coverage is 86.62%. Comparing base (261b0fb) to head (33630d6).

Files	Patch %	Lines
src/openoperators.jl	80.19%	20 Missing :warning:
src/acset.jl	0.00%	1 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #37 +/- ## ========================================== + Coverage 85.15% 86.62% +1.47% ========================================== Files 12 13 +1 Lines 788 890 +102 ========================================== + Hits 671 771 +100 - Misses 117 119 +2 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

lukem12345 commented 5 months ago

I'll stick a pin in 03aa7ce as a complete version of this feature in an imperative/ database-manipulation manner.

I'll continue commits to this branch from the operadic composition approach.

jpfairbanks commented 5 months ago

Let's keep the other approach in separate files so that we can compare and contrast the changes. I'm thinking the repurposing of oapply will be much more maintainable long term.

lukem12345 commented 5 months ago

Record of discussion with @GeorgeR227 :

Take care that the RHS Decapodes are fully expanded
Reformat (use newlines) in replace-operator functions
Merge the "more imperative" code (after review process) in 03aa7ce. (It is more declarative than existing code currently in Decapodes.)
Decide whether we should export the single-operator replacement functions.
Even the oapply approach will need some imperativity on top of it, either creating RHS Decapodes on-the-fly with the names of the src/tgt variables, or in the LHS, and the replaced operator still has to be explicitly removed.

lukem12345 commented 4 months ago

Default composition code has been cherry picked onto PR #52 . Upon that merging, I will reset this PR to 03aa7ce