Repository navigation
Repair precise public price-return expressions #107
Description
User workflow and ownership
Implement the accepted Engineer preparation/migration unit below. Preserve existing native defaults, dataframe contracts and independent use with supplied data. Data owns source adjustments/revision selection; Diagnostic owns the only Alphalens facade/cleaner and analytical workflows; Backtest owns simulation/accounting. Engineer produces native features, eligibility and outcomes, not competing engines.
Dependency: none. Start here. Recheck current main and owning records, execute the retained public
probe, and pin the actual failing source/environment before editing. Changes belong in
features/utils/arithmetic.py and reproduced downstream shared arithmetic only. Tests and
migration notes remain in the same numerical repair submission.
Preserve exact integer differences before division without signed/unsigned overflow, including
full-span differences. Cover simple/log/display percentage, lag, empty/single-row, invalid input,
zero denominators, Float32/64, null/NaN and grouped eager/lazy calls. Test declared Polars floor;
if a required type/API is unavailable, justify the narrow dependency-floor change and test it.
No early Float64 cast, whole-frame collect or Python row UDF should conceal lost ticks.
Proposed focused test file: tests/test_returns_precision.py. Verify with
uv run pytest tests/test_returns_precision.py -q, first failing on the selected producer base
for signed/unsigned tick loss and then passing on the repair. Ordinary controls must pass both.
Feed the actual output into a valid installed receiving return path; compounding is not another
Engineer engine. Inputs and row count remain unchanged and invalid declarations produce no output
write. Completion includes native artifact qualification, not just an adapter-specific workaround.
Premises
Input: price ticks
| Question | Answer | Kind |
|---|---|---|
| record | One exact typed price observation for one stable asset; return is decimal price change. | definitional |
| absent-record | An absent row supplies no price; an explicit null supplies an undefined quote, not a flat return. | definitional |
| ordering-spacing | Lag is ordered observations per asset; scheduled completeness is validated by the calling declared-grid workflow. | definitional |
| entity-identity | Explicit grouping keys separate histories, including interleaved input. | definitional |
| known-at | Primitive arithmetic does not assert causal availability; the caller's supplied prices and knowledge contract determine it. | definitional |
| institution | Exact integer arithmetic and declared simple/log definitions govern; no provider or inferred exchange rule. | definitional |
Claims
| Id | Claim | How it could be false |
|---|---|---|
| precise-native-return | Public eager/lazy integer returns retain each representable nonzero tick without overflow. | Adjacent high integers collapse to zero or full-span subtraction wraps. |
| return-api-stable | Ordinary/missing/domain/lag behavior and native shape remain compatible. | An overflow repair drops rows, collects a lazy plan or changes float/null behavior. |
Closes: price-ticks.record, price-ticks.absent-record, price-ticks.ordering-spacing, price-ticks.entity-identity, price-ticks.known-at, price-ticks.institution, precise-native-return, return-api-stable
Intake and completion
- Reproduce public producer defects before repairs and retain exact failing/passing revisions.
- Complete the methods, negative/error cases, runnable tests and migration route above.
- Verify actual installed outputs and the applicable final-revision quality/platform gates.
- Record producer/consumer dependencies separately from standalone qualification.
Plan reference: accepted Engineer preparation and Reloaded migration plan, unit 1 in milestone unreleased - Engineer preparation and Reloaded migration.
AI assistance: Codex prepared this intake projection from the accepted maintainer plan; implementation and independent test evidence are still required.
Activity
feat(labeling): produce native keyed multi-horizon outcomes (#114)
feat(features): port finite Pipeline factors and causal selection wor…