Milestone: Configurable TTL Cascade - #283
Merged
Merged
Conversation
…ble TTL cascade across ForeignKey graphs (milestone, phases 1-4)" This reverts commit 48e4adc.
…): CR-01 replay EVALSHA-only on NOSCRIPT, converge recovery paths execute_pipeline_with_noscript_recovery replayed the FULL command stack on NOSCRIPT. In a transactional MULTI/EXEC a NOSCRIPT surfaces at EXEC time after the non-EVALSHA commands already committed (Redis does not roll back a transaction on a mid-execution command error), so a full replay double-applies non-idempotent native ops (JSON.NUMINCRBY, JSON.ARRAPPEND, SF ops). This affected every write routed through ensure_pipeline/pipeline_with_execution, not just cascade, and only on a script-cache flush/failover. - context.py: replay only the EVALSHA entries (matching the already-correct _apipeline pattern); add an ignore_redis_error param so the two write paths share one recovery implementation (WR-03, IN-01). - base.py: _apipeline now delegates to the shared helper; drop the duplicated backup/replay block. Also document the non-positive-ttl root-delete asymmetry in aset_ttl(cascade=True) (IN-04) and restore valid indentation on contains_sf_field (stray uncommitted corruption). - test_context.py: rewrite the full-stack-replay test to assert EVALSHA-only replay (it had locked in the buggy behavior). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…): make Meta.ttl/cascade_ttl freeze lifecycle robust - init.py (WR-01): wrap the unfreeze -> configure -> bake -> refreeze sequence in try/finally so a failed init (e.g. validate_cascade_ttl_targets on a mis-configured graph) still refreezes every model instead of leaving them unfrozen with silently-mutable, half-baked Meta.ttl/cascade_ttl. - init.py (WR-02): teardown_rapyer now clears _ttl_frozen so a torn-down model doesn't leak MetaTtlFrozenError into a later init-less path. - config.py (WR-04): freeze cascade_ttl too, not just ttl — both are baked into the per-class Lua plan / _has_cascade gate, so mutating cascade_ttl post-freeze would silently desync the runtime cascade from the baked plan. - cascade/conftest.py: autouse fixture resets _ttl_frozen around each cascade test so the process-global freeze can't make tests order-dependent (WR-02). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…): harden lua-literal escaping and correct dead-branch comment - registry.py (IN-03): _lua_literal now also escapes newlines/CR so a stray control char in an injected literal yields valid Lua at SCRIPT LOAD instead of a silently broken script body. - apply.lua (IN-02): correct the recurse=false comment — the branch is a not-yet-exercised seam (every emitted edge has recurse=true), and a non-recursing target can still be reached via its override edges since next_hop ignores budget for overrides. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…): freeze the whole Meta config after init, not just ttl Per review: the freeze should cover all config, not only ttl. Rename the _ttl_frozen flag to _frozen, block every public Meta field in __setattr__ once frozen (private attrs stay writable so init/teardown can toggle it), and rename MetaTtlFrozenError -> MetaFrozenError. A global autouse test fixture unfreezes models around each test so the process-global freeze can't leak across tests. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…): refresh_ttl always uses the cascade script Per review: drop the _has_cascade branch in refresh_ttl and always run the cascade EVALSHA. With no outgoing edges it simply re-arms the model's own keys, so every TTL refresh goes through one path. Tests updated to assert the script call instead of the per-key EXPIRE loop. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…scade): planner dataclasses, contains_sf_field, shorter docstrings Per review, in rapyer/cascade/planner.py: - _classify_edge returns an EdgeClassification dataclass instead of a tuple. - _static_walk_special_suffixes uses contains_sf_field() instead of a manual hasattr(_special_field_names) check. - multi-line docstrings start on their own line and are trimmed to short summaries; drop -> None return annotations. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…de): remove internal workflow notation from comments Per review: drop the internal review/plan tags (D-0x, WR-0x, IN-0x, etc.) and RESEARCH.md/Pitfall references from comments and docstrings across the cascade code and tests; keep the actual explanations. Also reword multi-line docstrings to start on their own line. Renamed a few tests that embedded those tags. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
📝 WalkthroughWalkthroughTTL cascade support is added across configuration, cascade planning, atomic Lua execution, model TTL APIs, pipeline recovery, public exports, documentation, and unit/integration tests. Cascade traversal supports depth limits, special fields, dangling counts, cycles, shared nodes, and per-child TTLs. ChangesTTL Cascade
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant Model
participant CascadeScript
participant Redis
Model->>CascadeScript: Refresh root TTL with cascade
CascadeScript->>Redis: Read JSON foreign-key references
CascadeScript->>Redis: EXPIRE root, child, and special-field keys
Redis-->>Model: Return dangling child and special counts
Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Coverage reportTotal coverage: 99% Full report |
Merging this PR will degrade performance by 11.26%
Warning Please fix the performance issues or acknowledge them on CodSpeed. Performance Changes
Tip Investigate this regression by commenting Comparing 
|
yedidyakfir
commented
Jul 14, 2026
…on cascade script - cascade unit fixtures re-establish class-declared Meta.cascade_ttl (a prior init_rapyer() authoritatively resets it to None); restore on teardown - two integration tests save before clearing/flushing scripts, so the expected error fires inside pytest.raises rather than during asave() Arrange (asave now always runs the cascade Lua script via refresh_ttl) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (5)
tests/unit/test_context.py (1)
1-104: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winMissing coverage for the
ResponseErrorswallow/re-raise branch.
execute_pipeline_with_noscript_recoveryalso swallows non-NOSCRIPTResponseErrorwhenignore_redis_error=Trueand re-raises otherwise (per its docstring inrapyer/context.py), but no test here exercises either sub-path.♻️ Suggested additional tests
`@pytest.mark.asyncio` async def test_execute_pipeline_with_noscript_recovery_swallows_response_error_when_ignored(): pipe = _make_pipe( command_stack=[(("JSON.SET", "k", "$", "{}"), {})], execute_side_effect=ResponseError("boom"), ) result = await execute_pipeline_with_noscript_recovery( pipe, MagicMock(), ignore_redis_error=True ) assert result == [] `@pytest.mark.asyncio` async def test_execute_pipeline_with_noscript_recovery_reraises_response_error_by_default(): pipe = _make_pipe( command_stack=[(("JSON.SET", "k", "$", "{}"), {})], execute_side_effect=ResponseError("boom"), ) with pytest.raises(ResponseError): await execute_pipeline_with_noscript_recovery(pipe, MagicMock())As per path instructions, "Focus on test coverage completeness and edge cases" for
tests/**.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/unit/test_context.py` around lines 1 - 104, Add coverage for the ResponseError handling in execute_pipeline_with_noscript_recovery: import ResponseError and add async tests verifying a non-NOSCRIPT error returns an empty list when ignore_redis_error=True and is re-raised with the default setting. Use _make_pipe with a JSON.SET command and ResponseError side effect, while preserving existing tests.Source: Path instructions
rapyer/utils/annotation.py (1)
116-123: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueConsider reusing
extract_annotationto simplifyhas_annotation.To adhere to the DRY principle, you can update
has_annotation(lines 104-113) to simply check ifextract_annotationreturns a non-Nonevalue.♻️ Proposed refactor for `has_annotation`
def has_annotation(field: Any, annotation_type: Any) -> bool: return extract_annotation(field, annotation_type) is not None🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rapyer/utils/annotation.py` around lines 116 - 123, Update has_annotation to delegate to extract_annotation and return whether the result is non-None. Remove its duplicated annotation-origin and metadata traversal while preserving the existing boolean behavior.rapyer/config.py (1)
80-80: 🩺 Stability & Availability | 🔵 Trivial | 💤 Low valueUse
getattrto safely access_frozenduring object initialization or deserialization.While Pydantic usually populates
__pydantic_private__(wherePrivateAttrs are stored) directly during__init__without triggering__setattr__, certain object lifecycle states (such as unpickling or deep-copying) might invoke__setattr__before the private attributes dictionary is fully initialized, which would cause anAttributeError.Consider using
getattrto ensure robust attribute access.💡 Proposed fix
- if self._frozen and not name.startswith("_"): + if getattr(self, "_frozen", False) and not name.startswith("_"):🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rapyer/config.py` at line 80, Update the __setattr__ logic in the configuration model around the _frozen check to access _frozen safely with getattr and a false default. Preserve the existing name.startswith("_") condition and assignment behavior while preventing initialization, deserialization, or deep-copy lifecycle states from raising AttributeError.rapyer/scripts/lua/cascade/apply.lua (1)
244-275: 🚀 Performance & Scalability | 🔵 TrivialOperational note: unbounded cascades run as one long, blocking Lua script.
The full read-walk (all
JSON.GETs) plus theEXPIREwrite phase execute inside a single atomic EVALSHA, which blocks Redis's single-threaded command loop for its entire duration. With an unboundedCascadeTTL(no depth) or a deepoverridechain, a large/wide reachable subtree turns oneaset_ttl/refresh_ttlcall into a long-running script that stalls every other client on that Redis instance. This is an inherent tradeoff of the atomic-Lua design, not a defect — worth documenting as guidance to boundCascadeTTL(depth=...)on graphs that could grow large, and to monitor Lua execution time in production.Also applies to: 287-299
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rapyer/scripts/lua/cascade/apply.lua` around lines 244 - 275, Document guidance near plan_refresh_keys and the related refresh flow that unbounded or deeply overridden cascades execute all JSON reads and expiration writes in one blocking atomic Lua script. Recommend bounded CascadeTTL(depth=...) for potentially large or wide graphs and monitoring Lua execution time in production; do not alter the traversal or atomic execution behavior.rapyer/context.py (1)
48-99: 🎯 Functional Correctness | 🔵 Trivial | 🏗️ Heavy liftNOSCRIPT recovery path silently drops non-EVALSHA results; also relies on an internal Pipeline attribute.
Two related concerns on this helper:
commands_backup = list(pipe.command_stack)(line 69) reads redis-py's internalPipeline.command_stack— not documented public API. It has changed shape/behavior across redis-py releases for other pipeline variants (e.g.redis/redis-py#3703forClusterPipeline), so a future redis-py upgrade could silently break this without a semver signal.- On recovery, the function replays and returns only the
EVALSHAentries (lines 85-92), dropping any other commands originally queued in the same pipe from the returned list. This contradicts the docstring's claim that the success path "returns pipe.execute()'s result unchanged" — the recovery path returns a different, filtered list. Today every caller that queues EVALSHA-plus-other-commands in one pipe (ensure_pipeline's/pipeline_with_execution's exits,_apipeline) discards the return value, and the one caller that inspects it (aset_ttlin base.py) only ever uses a dedicated single-EVALSHA pipe — so this isn't live today, but it's a footgun for the next caller that mixes EVALSHA with other pipe commands and trusts positional results after a NOSCRIPT event.Please confirm the target redis-py version continues to expose
command_stackin this shape, and whether any call path (e.g. viamark_actions/refresh_ttl(can_use_pipeline=True)) nests the cascade EVALSHA into a pipe alongside other writes while consuming this function's return value.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@rapyer/context.py` around lines 48 - 99, Update execute_pipeline_with_noscript_recovery to avoid depending on redis-py’s internal command_stack shape: verify the supported redis-py version and capture queued command metadata through a stable, explicit mechanism. Audit callers such as mark_actions, refresh_ttl, ensure_pipeline, pipeline_with_execution, _apipeline, and aset_ttl, then make NOSCRIPT recovery preserve the original command-result positions instead of returning only replayed EVALSHA results. Keep dedicated EVALSHA callers’ behavior unchanged and document or enforce the resulting recovery contract for mixed pipelines.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@rapyer/cascade/planner.py`:
- Around line 178-184: Update the field-class resolution in the loop over
model_cls._contain_sf before calling safe_issubclass: unwrap Optional/Annotated
annotations using the existing strip_optional and get_origin logic, matching
_unwrap_nested_model_cls, then perform the AtomicRedisModel check on the
unwrapped nested model class.
In `@tests/integration/foreign_keys/test_cascade_ttl_apply.py`:
- Around line 22-25: Update the TTL assertion in the test using
SCRIPT_FLUSH_ROOT_TTL_SECONDS so it verifies the resulting parent TTL is bounded
by the explicit 120-second value, rather than merely being positive. Ensure the
test setup persists parent and child as needed before invoking the script flush,
while preserving the existing distinction from CASCADE_FIXTURE_TTL_SECONDS and
the current assertion flow.
---
Nitpick comments:
In `@rapyer/config.py`:
- Line 80: Update the __setattr__ logic in the configuration model around the
_frozen check to access _frozen safely with getattr and a false default.
Preserve the existing name.startswith("_") condition and assignment behavior
while preventing initialization, deserialization, or deep-copy lifecycle states
from raising AttributeError.
In `@rapyer/context.py`:
- Around line 48-99: Update execute_pipeline_with_noscript_recovery to avoid
depending on redis-py’s internal command_stack shape: verify the supported
redis-py version and capture queued command metadata through a stable, explicit
mechanism. Audit callers such as mark_actions, refresh_ttl, ensure_pipeline,
pipeline_with_execution, _apipeline, and aset_ttl, then make NOSCRIPT recovery
preserve the original command-result positions instead of returning only
replayed EVALSHA results. Keep dedicated EVALSHA callers’ behavior unchanged and
document or enforce the resulting recovery contract for mixed pipelines.
In `@rapyer/scripts/lua/cascade/apply.lua`:
- Around line 244-275: Document guidance near plan_refresh_keys and the related
refresh flow that unbounded or deeply overridden cascades execute all JSON reads
and expiration writes in one blocking atomic Lua script. Recommend bounded
CascadeTTL(depth=...) for potentially large or wide graphs and monitoring Lua
execution time in production; do not alter the traversal or atomic execution
behavior.
In `@rapyer/utils/annotation.py`:
- Around line 116-123: Update has_annotation to delegate to extract_annotation
and return whether the result is non-None. Remove its duplicated
annotation-origin and metadata traversal while preserving the existing boolean
behavior.
In `@tests/unit/test_context.py`:
- Around line 1-104: Add coverage for the ResponseError handling in
execute_pipeline_with_noscript_recovery: import ResponseError and add async
tests verifying a non-NOSCRIPT error returns an empty list when
ignore_redis_error=True and is re-raised with the default setting. Use
_make_pipe with a JSON.SET command and ResponseError side effect, while
preserving existing tests.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 8b870c71-cf05-45da-9f3c-905ff49267ba
📒 Files selected for processing (47)
docs/documentation/special-fields/foreign-keys.mddocs/documentation/special-fields/ttl-cascade.mdmkdocs.ymlrapyer/__init__.pyrapyer/base.pyrapyer/cascade/__init__.pyrapyer/cascade/planner.pyrapyer/cascade/spec.pyrapyer/cascade/ttl.pyrapyer/config.pyrapyer/context.pyrapyer/errors/__init__.pyrapyer/errors/cascade.pyrapyer/init.pyrapyer/result.pyrapyer/scripts/constants.pyrapyer/scripts/lua/cascade/__init__.pyrapyer/scripts/lua/cascade/apply.luarapyer/scripts/registry.pyrapyer/utils/annotation.pytests/conftest.pytests/integration/foreign_keys/conftest.pytests/integration/foreign_keys/test_cascade_action_boundary.pytests/integration/foreign_keys/test_cascade_concurrent_mutation.pytests/integration/foreign_keys/test_cascade_graph_shapes.pytests/integration/foreign_keys/test_cascade_ttl_apply.pytests/integration/lst/test_redis_list_remove_range.pytests/integration/pipeline/test_pipeline_noscript_recovery.pytests/models/cascade_types.pytests/unit/cascade/__init__.pytests/unit/cascade/conftest.pytests/unit/cascade/test_aset_ttl_cascade_flag.pytests/unit/cascade/test_cascade_action_boundary.pytests/unit/cascade/test_cascade_apply_lua.pytests/unit/cascade/test_cascade_apply_lua_syntax.pytests/unit/cascade/test_cascade_classification.pytests/unit/cascade/test_cascade_plan_injection.pytests/unit/cascade/test_cascade_plan_table.pytests/unit/cascade/test_cascade_ttl_config.pytests/unit/cascade/test_cascade_ttl_required_validation.pytests/unit/cascade/test_extract_annotation.pytests/unit/cascade/test_init_rapyer_cascade_ttl.pytests/unit/cascade/test_meta_ttl_freeze.pytests/unit/cascade/test_refresh_ttl_cascade_branch.pytests/unit/test_context.pytests/unit/test_init_rapyer.pytests/unit/test_refresh_ttl_if_needed.py
…est comments (PR #283 review) - Rename CascadeEdge fields (collection/recurse/ttl/special/override -> is_collection/recurse_into_target/refresh_target_ttl/ refresh_target_special_keys/resets_depth_budget) across planner.py, apply.lua, and every test that inspects them; document the always-True flags as forward-looking per-edge hooks. - Rename RedisConfig._frozen -> _meta_locked across config.py, init.py, and tests. - Fix multi-line docstrings that started text on the opening `"""` line. - Unwrap Optional/generic annotations before the subclass check in _static_walk_special_suffixes (CodeRabbit #14). - Tighten the SCRIPT_FLUSH_ROOT_TTL_SECONDS assertion in test_cascade_ttl_apply.py to prove the explicit root ttl was applied, not just that some positive ttl survived (CodeRabbit #15). - Reformat inline Arrange/Act/Assert comment markers onto their own header line across the new cascade tests (CodeRabbit #12/#13). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
… TTL feature (PR #283 #10) - context.py: remove execute_pipeline_with_noscript_recovery and its now-unused imports; ensure_pipeline/pipeline_with_execution revert to a bare pipe.execute() (develop's behavior). - base.py: restore develop's self-contained inline NOSCRIPT recovery inside _apipeline (EVALSHA-only replay + PersistentNoScriptError on second failure); aset_ttl's standalone execute is now bare, matching the TTL-refresh paths' new (documented) lack of self-heal. - Update/remove tests that asserted the now-removed generic recovery seam; test_pipeline_noscript_recovery.py's two script-flush tests now flush after the establishing asave() so only the explicit apipeline() block (backed by _apipeline) needs to recover. - Track extending self-heal to the TTL-refresh paths as a follow-up (see .planning/quick/260714-l0p-fix-9-failing-tests-on-pr-283-cascade-tt/NOSCRIPT-ISSUE.md). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
yedidyakfir
commented
Jul 14, 2026
- Remove the _has_cascade ClassVar from AtomicRedisModel and the marking loop in init_rapyer -- it was written but never read since aset_ttl/refresh_ttl unified onto the cascade Lua script, which is the sole source of cascade-traversal truth via the plan table baked in at register_scripts time. - build_cascade_plan/validate_cascade_ttl_targets still run in init_rapyer for fail-fast config validation. - Strip the matching _has_cascade stash/restore scaffolding from six test files; pure pass-through wrapper fixtures are removed in favor of the base fixture they wrapped. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…iew) Condense refresh_ttl's and aset_ttl's over-long cascade-ARGV comment blocks down to their essential why -- comment text only, zero logic change (should_execute=False / manual pipe.execute() pattern, ARGV order, and CascadeResult construction are untouched). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
yedidyakfir
commented
Jul 15, 2026
…ry.lua Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…disConfig - Add cascade_function_name field, init-baked (None on fakeredis) - Exempt it from the _meta_locked freeze guard so arun_fcall self-heal can rewrite it Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Delete _CASCADE_FUNCTION_NAME module global and get_cascade_function_name - register_cascade_function returns the plan-hashed name instead of writing a global - run_fcall takes function_name explicitly; arun_fcall reads/refreshes it from redis_config - handle_missing_function assigns the refreshed name onto redis_config - Drop get_cascade_function_name from scripts package exports Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ad it to FCALL - init_rapyer assigns register_cascade_function's return onto every Meta post-freeze - Both base.py run_fcall sites (refresh_ttl, aset_ttl) pass self.Meta.cascade_function_name Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…sertions and conftests - Unit assertions insert Meta.cascade_function_name as run_fcall's new second positional arg - Integration conftests capture register_cascade_function's return onto Meta and restore in teardown Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Delete unused extract_annotation from rapyer/utils/annotation.py - Delete its sole test file tests/unit/cascade/test_extract_annotation.py - field_with_flag (production superseder) and has_annotation untouched Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ecute sites - Add shared aexecute_pipeline_with_cascade_self_heal + aretry_fcall_after_missing_function to registry.py - ensure_pipeline/pipeline_with_execution route execute through the self-heal wrapper (lazy import, documented cycle) - aset_ttl routes bare execute through the self-heal wrapper - _apipeline replays FCALL on function-not-found, reusing aretry_fcall_after_missing_function - Rewrite config.py freeze-exempt comment; drop resolved issue #284 notes - Retry rewrites only the function-name slot; single retry then PersistentCascadeFunctionError - fakeredis EXPIRE branch and single-FCALL atomicity preserved Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…call in test helpers - Delete arun_fcall from registry.py (import + __all__ in scripts/__init__.py) - handle_missing_function and PersistentCascadeFunctionError retained (used by self-heal) - 3 integration _apply_cascade helpers call real_redis_client.fcall directly - Production self-heal now covers what arun_fcall did (Task 2) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…284, RED) - After FUNCTION FLUSH, aset_ttl(cascade=True) and refresh_ttl must reload the cascade function and still refresh the reachable subtree - Currently fails: redis-py's async pipeline masks the function-not-found message, so the helper's string match never fires inside a pipeline Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…not error text (GREEN) - redis-py's async pipeline annotate_exception (client.py:1585) overwrites exception.args with a non-f-string literal, destroying the "Function not found" message; every production FCALL runs inside a pipeline, so string-matching never fired there (only the deleted direct-client arun_fcall saw the real text) - Add acascade_function_missing (FUNCTION LIST scan) + _pipeline_has_fcall guard - Self-heal only fires on real Redis when the pipeline enqueued an FCALL and the cascade function is genuinely absent; other ResponseErrors re-raise unchanged - Wire the registry-based detection into both the shared wrapper and _apipeline Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…drop cascade self-heal helpers - context.py: ensure_pipeline + pipeline_with_execution use bare pipe.execute(), remove lazy registry imports - base.py: aset_ttl uses pipe.execute(); _apipeline reverts to NOSCRIPT-only path (no FCALL-missing detection/replay) - registry.py: delete six self-heal helpers; imports drop ResponseError, PersistentCascadeFunctionError, cascade.planner Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…te self-heal test, fix stale comments - errors/cascade.py + __init__.py: remove PersistentCascadeFunctionError and its __all__ entry - delete tests/integration/foreign_keys/test_cascade_self_heal.py - config.py: cascade_function_name freeze-exemption comment now references init_rapyer(), not the removed self-heal path Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
yedidyakfir
changed the title
6 tasks
This was referenced Jul 21, 2026
Merged
yedidyakfir
pushed a commit
that referenced
this pull request
Aug 13, 2026
…eased entry since v1.3.4 Only #288 had an entry; five merged items were missing. - [1.3.6] new: SF-held reference cascade (#289, Added) and multi-class union / polymorphic-base FK reach (#290, Fixed) - [1.3.5] backfilled: the configurable TTL cascade feature itself (#283), the init_rapyer client-rebind fix (#276), runtime CPU optimizations (#263) Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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
2 participants
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.
The complete Configurable TTL Cascade milestone, re-applied on top of the reverted
develop(after #279) with every review comment from #277 addressed — plus the full evolution since. This is the PR to review the whole diff in and merge cascade back intodevelop.Summary
Milestone: Configurable TTL Cascade — the first slice of a configurable cascade framework for
rapyer. Setting a parent aggregate's TTL performs a cascading refresh across its cascade-enabledForeignKey-referenced children — each child re-armed to its ownMeta.ttl— applied atomically and server-side at set-time. Cascade is opt-in and disabled-by-default, so existing projects are byte-for-byte unaffected.Status: All 4 phases complete and verified; full CI matrix green.
Phases (all complete)
CascadeTTL; FK-graph field classification feeding the cascade plan;CascadeSpecEXT-01 seam for future delete/save cascade.Meta.ttlin a single atomic op, with a per-class plan baked in.Evolution since the initial milestone (quick-task refinements)
EVALSHAscript to aFUNCTION LOAD/FCALLlibrary (library.lua): the plan is baked into the library and decoded once (memoized upvalue;cjsonis unavailable at load scope), so per-call cost is ~0. Library + function names carry a plan-hash for server-global isolation.FUNCTIONsupport, so it falls back to a root-ownall_keysEXPIREloop — preservingMeta.ttl/refresh_ttlbehavior. Cascade traversal tests are gated to real Redis 7+.RedisConfig.cascade_function_name(init-baked, freeze-exempt).extract_annotationand the test-onlyarun_fcallwrapper.Verification
black --check+ruff checkclean;mypy(3.10–3.13) green.Note on the CodSpeed check
The only red check is CodSpeed Performance Analysis.
developcurrently has no cascade at all (reverted in #279), so any cascade feature registers as a "regression" against a cascade-free baseline — this is the feature's inherent cost, not a defect. Against the last cascade-bearing benchmark baseline, the Functions rewrite is 11–46% faster with 0 regressions.Housekeeping
Supersedes #278 (which showed only the fixes because its branch shared the milestone commit as a merge-base with develop). #278 can be closed.
🤖 Generated with Claude Code