Feature collection: parser QoL#277
Open
liamhuber wants to merge 31 commits into
Open
Conversation
Adds a source-node recipe for emitting fixed JSON-serializable values: RecipeElementType.CONSTANT enum member, the ConstantRecipe pydantic model (with a strict before-validator rejecting tuples/sets/non-str keys), its registration in the discriminated union, and public exports.
Literal call arguments (e.g. mul(0.5, mv_2)) previously failed to parse because consume_call_arguments required every argument to be an ast.Name. Non-Name arguments are now literal_eval'd and injected as a ConstantRecipe source node wired to the consuming port; SymbolScope.consume_source records the resulting sibling edge without polluting the symbol table. The ConstantRecipe model's mode="before" validator remains the single source of truth for the JSON-compatibility invariant. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…oat constants; cleanups Address code-review findings from the constant-node-type feature: - case_helpers.parse_case discarded scope_copy.edges, so a literal argument in an if/elif/while condition injected an orphaned constant node and left the condition's input sourceless downstream. consume_call_arguments now takes allow_constants=False for conditions and raises TypeError up front instead. - ConstantRecipe accepted non-finite floats (nan/inf/-inf), which are not valid JSON and cannot round-trip through ast.literal_eval; _strict_json now rejects them. - Formatting and test-coverage cleanups: black on test_parsing_constants.py, removed redundant in-method imports in test_retrospective_wfms.py, restored partial-consumption-before-error coverage in test_parser_helpers.py, and a clarifying comment in compiler/statements.py. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ition A constant feeding a workflow output, or sitting bare as a flow-control body, can't be inlined (`return 5` doesn't re-parse; a bare ConstantRecipe crashes _emit_single_node_body on node.reference). Add a materialize path: emit_body gets a ConstantRecipe branch ahead of _emit_single_node_body, and emit_workflow_body tracks constants whose output feeds a workflow output edge, emitting `sym = repr(value)` for those instead of inlining. Constants feeding only call arguments still inline unchanged. The narrowing in emit_body's new isinstance branch resolves the host-only mypy arg-type/union-attr errors at statements.py's _emit_single_node_body call site (mypy segfaults in this sandbox, so this is asserted structurally, not re-run here).
body.value is expr|None for AnnAssign; short-circuit on None so a valueless annotation still falls through to the existing error path. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Route each literal condition argument to a synthetic flow-control input port fed by a constant peer injected into the enclosing workflow, replacing the allow_constants=False rejection. Parser-only; the compiler already renders the peer-fed condition argument inline. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ions Adds TestConstantsInConditions to test_compiler.py proving if/elif/while literal conditions execute correctly and round-trip exactly (including multiple-literals-in-one-condition and nested cases), plus a determinism-pin test and a non-JSON-literal-raises test in test_if_parser.py. Confirms allow_constants is fully gone (grep). Also fixes a stale comment in statements.py: a constant peer feeding a flow-control input is normal now (Part 2), not an unreachable hand-built shape as the old prose claimed; the conflict `if` below it remains correctly unreachable in parser-produced recipes for a different reason (such a peer is never also a workflow-output source). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
And keep it in a classvar Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
With safeguards against interactions in flow controls: no reassigning to accumulators; no conditionals leaking input sources. Co-authored-by: Claude <noreply@anthropic.com> Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
And update error message Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
And plug a dangerous edge case where the signature comes from `__new__` by failing cleanly. Co-authored-by: Claude <noreply@anthropic.com> Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
When decorating classes Co-authored-by: Claude <noreply@anthropic.com> Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
That stacks @atomic and @DataClass Co-authored-by: Claude <noreply@anthropic.com> Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Signed-off-by: Liam Huber <liamhuber@greyhavensolutions.com>
Parsing constants
|
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #277 +/- ##
==========================================
Coverage 100.00% 100.00%
==========================================
Files 35 38 +3
Lines 2818 3037 +219
==========================================
+ Hits 2818 3037 +219 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Parse variable reassignment
Refactor multiple dispatch
Parse classes as atomic recipes
Shortcut dataclass decorator
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is a landing pad for a set of PRs focused around improving quality of life in parsing decorated objects. It's driven by annoyances I encountered formulating @ligerzero-ai's
pyiron_workflow_atomisticsnodes intoflowreprecipes -- I never ran into anythingflowrepcouldn't do, but there were a bunch of places that I wish it made things easier for me.The only change to the actual model side, and the only change I anticipate right now, was adding the "constant"-type recipe in #270. The changes really are to the parsing/compiling.
I'm going to try and keep actual changes fairly focused into individual PRs, but when there's more than a small handful of these stacked on top of each other, propagating requested changes from the base on up the stack is just super annoying. So different changes will be digestible and reviewable in their own PRs, but requested changes will be patched in here in the aggregator stage to avoid that propagation.
"dataclass"unpack mode without loss of usablity