Files
mt5cli/tests
Daichi Narushima b878a61c07 fix: re-verify normalized volume margin in calculate_volume_by_margin (#46)
* fix: re-verify normalized volume margin before returning from calculate_volume_by_margin

For CFDs, index products, and tiered-margin instruments, the initial
min-lot margin estimate can be optimistic; the normalized stepped volume
may require more margin than available_margin.  After computing the
normalized volume, step down by volume_step until order_calc_margin
confirms affordability, or return 0.0 if no step is affordable.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* test: fix ruff line-length violations in calculate_volume_by_margin tests

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* fix: use integer step index and add actual>0 guard in calculate_volume_by_margin

Replace float-subtraction loop with integer step index to eliminate
accumulation rounding error and add `actual > 0` guard so a broker
returning zero/negative margin is never accepted as affordable.
Inline `capped` to keep local-variable count within Ruff PLR0914 limit.
Update docstring to reflect re-verification behaviour and 0.0 fallback.
Tighten test assertion from `volume > 0` to the symbol's valid range.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* Bump version to v0.9.1

* perf: replace linear step-down scan with binary search in calculate_volume_by_margin

Resolves the P2 review finding: the previous O(n) loop called
order_calc_margin once per volume step, making sizing appear hung for
symbols with a large step range or small volume_step.

Binary search over the integer step index finds the largest affordable
step in O(log n) IPC calls (≈17 for a 99 999-step range vs up to 99 999
in the worst case). Monotonicity of broker margin with volume is assumed,
which holds for standard linear margin schedules.

To stay within the PLR0914 local-variable limit the steps variable is
inlined into hi and the tick temporary is eliminated by accessing the
snapshot dict directly. Error messages still go via msg to satisfy EM102.

Two existing tests are updated to match the binary-search call sequence.
A new regression test (volume_min=0.01, volume_max=1000.0) configures
a tiered-margin mock with its threshold at step 50000 and asserts that
the total order_calc_margin call count does not exceed 25.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* chore: remove obsolete TC003 per-file-ignore for history.py

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

---------

Co-authored-by: agent <agent@localhost>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-23 04:40:00 +09:00
..