Sentinel/Azimuth/bars/test_renko.py
| Family | Azimuth (Python) |
| Version | — |
| Size | 594 lines |
| Scope | public — ships in sentinel-suite |
| Documented by | SENTINEL_AZIMUTH_SPEC |
Rendered from the published copy in
sentinel-suite/src/, not the author’s private tree — so this page describes the file you actually have.
Tests for the Renko port and the `bars` package plumbing.
C:\\ntbv\\Scripts\\python.exe -m pytest bars\\test_renko.py -q
C:\\ntbv\\Scripts\\python.exe bars\\test_renko.py # same proofs, no pytest
Two kinds of test, and both are needed. The HAND WALK is the only thing that pins the
five behaviours the C# actually has (wickless bricks, the forming bar's stamp and
volume, row-less gap bricks, the session doji, the two-brick reversal) -- an invariant
suite would happily pass on a wrong-but-self-consistent Renko. The INVARIANTS over the
real `GC 02-26` tape are what say the port survives 1.5 million rows of real data.
⚠ Neither is the gate. Passing these does NOT mean the port matches NinjaTrader; only
`bars.gate` can say that. It HAS now been run (2026-08-05, all 17 `GC 02-26` sessions
against `20260805T015237__GC__11v1x1.jsonl`) and it does NOT yet pass -- see README.
Everything it caught that this file could have caught is now pinned below.