PyneCore v6.8.0 released — most of the ta.* rolling machines, the math.sum accumulator,…

August 4, 2026

Features

Pine’s tolerant float comparison (absolute 1e-10 at every magnitude) now applies to comparison operators in Pyne code and to the builtins measured to compare tolerantly
OHLCV v2 stores the symbol’s tick grid in the header; delta-decoded prices snap back onto it, so 65399.99 no longer reads as 65399.990000000224
Atomic file publication and file locking now work on Windows

Fixes

math.sum, ta.sma, ta.ema, ta.rma, ta.wma, ta.hma, ta.variance, ta.stdev, ta.correlation, ta.linreg, ta.mfi, ta.stoch, ta.rci, ta.percentile_*, ta.change and math.avg are bit-exact with TradingView; ta.atr, ta.rsi, ta.macd, ta.tsi and ta.dmi follow
str.tostring and str.format print TradingView’s digits
ta.cross no longer short-circuits, so a crossunder right after a crossover is not missed
A limit order the bar gaps past no longer takes market slippage (#74)
strategy.position_avg_price re-weights correctly when pyramiding, and the cash_per_order commission reaches the closed trade’s own profit
Volume from a v2 feed is used as served instead of being truncated by legacy float32 clean-up, which had diverged ta.vwap

Performance

ta.wma, ta.linreg and ta.dev about 5x cheaper per bar at length 500
ta.percentile_nearest_rank and ta.percentile_linear_interpolation 2.5x-100x faster
One ta.sma call: 1811 ns -> 686 ns