My chart knew how 2019 would end. The replay was still in March.
During work on XTester Studio, the yearly candle was already showing the full year's high and close. The monthly chart looked fine, which made the difference worth investigating.
The chart had loaded its history before the replay started and kept that window in memory. Starting the replay hid candles that belonged entirely to the future, but the current year's candle stayed there, complete. The simulation engine was cutting the data at the right date. The chart was displaying an old, unrestricted view.
The monthly chart had been opened after the replay began. That small difference in when a chart was opened explained why the two looked different.
The fix and regression tests are merged. Checking the behavior in a fresh running build is still on the list.
For anyone practicing on historical charts, this is an awkward kind of bug: a candle can sit at the correct date and still contain prices you shouldn't have seen yet.
During work on XTester Studio, the yearly candle was already showing the full year's high and close. The monthly chart looked fine, which made the difference worth investigating.
The chart had loaded its history before the replay started and kept that window in memory. Starting the replay hid candles that belonged entirely to the future, but the current year's candle stayed there, complete. The simulation engine was cutting the data at the right date. The chart was displaying an old, unrestricted view.
The monthly chart had been opened after the replay began. That small difference in when a chart was opened explained why the two looked different.
The fix and regression tests are merged. Checking the behavior in a fresh running build is still on the list.
For anyone practicing on historical charts, this is an awkward kind of bug: a candle can sit at the correct date and still contain prices you shouldn't have seen yet.
