average_price edges; pin current IOC/GTD behavior at the level (#79)#97
Conversation
…the level - average_price: tests for zero-executed (Ok(None)), exact small values, the f64 precision-loss case at 2^53+1 (finite, never NaN/Inf — documented intrinsic limitation), and a never-NaN/Inf sweep. - IOC de-facto pin: taker exceeding depth fills available and reports the remainder; the taker is never enqueued. Real IOC/FOK taker enforcement is gated on #65. - GTD/Day: test that match_order does NOT enforce maker expiry (a past-expiry Gtd maker still matches); added a match_order rustdoc note that TIF expiry is the caller's responsibility. Clarifying comments on the existing FOK/IOC maker tests. FOK all-or-nothing enforcement deferred to #65.
joaquinbejar
added
the
price-level
There was a problem hiding this comment.
Pull request overview
Adds targeted test coverage and documentation to pin current PriceLevel::match_order semantics around time-in-force boundaries (IOC/FOK/GTD) and to extend MatchResult::average_price() coverage for numeric edge cases, without changing matching behavior.
Changes:
- Adds
average_price()edge-case tests (zero-executed →Ok(None), precision-loss case, and finiteness sweep). - Adds new
PriceLeveltests to pin “taker exceeds depth” remainder behavior and to pin that maker GTD expiry is not enforced at this layer. - Updates
PriceLevel::match_orderrustdoc to explicitly state maker TIF expiry is caller/order-book responsibility, and clarifies existing IOC/FOK maker-TIF tests via comments.
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| src/price_level/tests/level.rs | Adds tests pinning taker-over-depth remainder behavior and maker GTD-expiry non-enforcement; clarifies existing IOC/FOK maker-TIF tests. |
| src/price_level/level.rs | Documents that maker TIF expiry is not enforced in match_order (caller responsibility). |
| src/execution/tests/match_result_trade.rs | Adds MatchResult::average_price() edge-case/precision/finite tests. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Codecov Report✅ All modified and coverable lines are covered by tests. ❌ Your project status has failed because the head coverage (72.71%) is below the target coverage (75.00%). You can increase the head coverage or adjust the target coverage.
... and 1 file with indirect coverage changes 🚀 New features to boost your workflow:
|
…fix comment The precision test compared against EXACT_INT_AVG as f64, which rounds the same way as the result (abs_err always 0), so it did not actually show the loss. Now converts the f64 average back to u128 and asserts it is short of the exact integer average by exactly 1. Also corrected a module comment that claimed happy-path coverage lived in the price-level tests.
|
Thanks, both fixed. The off-by-1 is now shown in integer space (the f64 result converted back to u128 is short by exactly 1), and the module comment points to the right coverage. |
Summary
Adds the non-gated parts of the TIF / expiry /
average_pricetest coverage:average_priceedge cases, a pin of the level's de-facto IOC behavior, and atest + doc note that time-in-force expiry is the caller's responsibility. The
parts that need taker-TIF enforcement (FOK all-or-nothing, true IOC discard) are
deferred to the gated #65.
Changes
average_priceedges (match_result_trade.rs): zero-executed →Ok(None);exact small values; the
f64precision-loss case; and a never-NaN/Infsweep over valid inputs.
level.rstests):test_match_order_taker_exceeds_depth_fills_available_and_reports_remainder—a larger-than-depth taker fills available, leaves
remaining_quantity() > 0,is_complete() == false, and is never enqueued.test_match_order_does_not_enforce_gtd_maker_expiry— a Gtd makerwith a past expiry still matches; added a
match_orderrustdoc note that TIFexpiry is not enforced at the level (caller/order-book responsibility).
test_match_fill_or_kill_order/test_match_immediate_or_cancel_order(maker TIF not enforced; matches likeGtc).
Technical decisions
average_priceprecision (no bug): forprice = 2^53 + 1, quantity = 1the exact average is
9_007_199_254_740_993but thef64division yields…992.0(off by 1) —2^53 + 1isn't representable in an f64 mantissa. Thematching path keeps exact integer math; only
average_priceanalytics drop tof64. NeverNaN/Inffor valid inputs. Tests assert closeness + finitenessand document the limitation.
match_ordertakes no takerTimeInForceand never enqueues the taker, so itcan't distinguish fill-or-kill from fill-what-you-can. The IOC test pins only
the observable behavior the engine already has.
Testing
make pre-pushcleancargo test: 380 (unit) + 1 (integration) + 9 (doctests) passed, 0 failed.cargo clippy --all-targets --all-features -- -D warnings,cargo fmt --all --check,cargo build --release: all clean.Checklist
average_pricezero / precision / NaN-guard edges coveredCloses #79