improvement(semantic/cfg): rework basic blocks.

[rzvxa]

I've replaced the BasicBlockElement with an Instruction type which would keep both the instruction kind and its associated AstNodeId.
I also removed the register scheme in the control flow in favor of a simpler approach using explicit enums.

#3381 (comment)

[graphite-app]

Your org has enabled the Graphite merge queue for merging into main

Add the label “merge” to the PR and Graphite will automatically add it to the merge queue when it’s ready to merge. Or use the label “hotfix” to add to the merge queue as a hot fix.

You must have a Graphite account and log in to Graphite in order to use the merge queue. Sign up using this link.

[rzvxa]

• feat(linter/eslint): add no_unreachable rule. #3238 : 2 dependent PRs (#3631 , #3650 )
• feat(semantic/cfg): propagate unreachable edges through subgraphs. #3648
• feat(semantic/cfg): add Condition instruction. #3567
• feat(semantic/cfg): add iteration instructions. #3566
• improvement(semantic/cfg): better control flow for switch statements. #3547
• improvement(semantic/cfg): rework error path. #3519
• improvement(semantic/cfg): better throw control flow. #3473
• improvement(linter/react): use CFG for detecting cyclic subgraphs in rules-of-hooks. #3454
• improvement(semantic/cfg): better CFG API #3472
• improvement(semantic/cfg): better break and continue flow. #3462
• improvement(semantic/cfg): better control flow for ForStatements. #3453
• improvement(semantic/cfg): better control flow for DoWhileStatements. #3452
• improvement(semantic/cfg): better control flow for WhileStatements. #3451
• improvement(semantic/cfg): rework basic blocks. #3381 👈
• fix(linter): memorize visited block id in neighbors_filtered_by_edge_weight #3407
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @rzvxa and the rest of your teammates on Graphite

[codspeed-hq]

CodSpeed Performance Report

Merging #3381 will degrade performances by 3.89%

_{Comparing 05-22-improvement_semantic_cfg_rework_basic_blocks (209a99d) with main (646b993)}

Summary

⚡ 1 improvements
❌ 1 regressions
✅ 20 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	main	05-22-improvement_semantic_cfg_rework_basic_blocks	Change
⚡	transformer[RadixUIAdoptionSection.jsx]	286.1 µs	267.7 µs	+6.9%
❌	transformer[pdf.mjs]	61 ms	63.5 ms	-3.89%

[mysteryven]

Look forward to this series of works. Looks like we can get AstNode directly by visiting CFG after this, Will it provide more features?

[rzvxa]

Look forward to this series of works. Looks like we can get AstNode directly by visiting CFG after this, Will it provide more features?

This PR will introduce Instruction to replace BasicBlockElement, They're really similar but since Instruction can be any statement, We would be able to check if a given statement in a BasicBlock comes exactly before or after a jump/return; On top of this, we can build better inference and a more ergonomic API that might mimic the path segment stacks more closely.

So there are 3 major points that I'm trying to achieve. Some will happen in this PR and some I will do shortly.

1. Associate a CFG instruction with its origin node if one exists(we have dummy blocks with no node so it is optional)
2. Be more specific on instruction types and express a jump as an instruction(in addition to its edge which already exists). This helps us infer whether a jump happens before or after another statement, It might be unnecessary but it's too soon for me to say for sure. Could be possible to provide this information with a better API instead of adding redundant information?
3. Keep track of the unwinding path and provide a way to follow the error path up the stack.

With all that said, I'm pretty much-adding features and refactoring stuff as I find shortcomings while implementing something. So no cleaver querying and/or super specific CFG visitors are going to come out of this; Unless somebody reports an actual use case for it.

[rzvxa]

Reviewing this stack is becoming more cumbersome by the minute, I'll mark some parts of it that I'm sure are correct for review.
Is there a kind soul out there who would put up with these changes and give it a spin?

I'll write a summary for each PR I mark as "ready".

[graphite-app]

Merge activity

• Jun 6, 1:40 AM EDT: Boshen added this pull request to the Graphite merge queue.
• Jun 6, 1:46 AM EDT: Boshen merged this pull request with the Graphite merge queue.