Repository navigation
Support TA-Lib 0.8.3 C library - #785
Merged
Merged
Conversation
wheels.yml no longer builds TA-Lib C, so its TALIB_C_VER was read only by tests.yml. Read the version from setup.py instead, so a bump cannot leave the shared-library test on the old release.
Output bit 128 is now a confirmation level (+/-200 on the bar that confirms an earlier pattern), not the old strength range. Output bits 0x4000 (display shift) and 0x8000 (weak pattern, +/-80) and function bits 0x2, 0x4 and 0x8 were dropped: output_flags could not tell that CDLENGULFING writes +/-80. Refs #126
A Function call that raised skipped the restore of the parameters and price names it was given, so they stuck and the next call used them: after abstract.SMA(d, timeperiod=0) failed, abstract.SMA(d) failed too. Restore them on every path. set_parameters also checks every value before writing any, so a rejected call leaves the function as it was. Refs #194
`make cython` from the repo root, with Cython 3.3.0 and numpy 2.5.4 in .venv (python3.12). Built from this file without Cython: 1714 passed, 1 skipped.
The generators added begidx to every integer output of a function whose name contains INDEX. TA-Lib C 0.8.3's ZIGZAG returns a bar index in outPivotIdx next to a trend direction in outTrend, so the shift is now chosen per output: every output of the *INDEX functions, as before, and any integer output whose C name ends in Idx. The generated code of MAXINDEX, MININDEX and MINMAXINDEX is unchanged.
… C 0.8.3 TA-Lib C 0.8.3 adds 36 functions (ALMA, ASI, BBW, CG, CHOP, CHOPTR, CKSP, CRSI, CTI, DOSC, EMV, FISHER, FRAMA, IBS, ICHIMOKU, KST, KSTEXT, KURTOSIS, KVO, MCGD, MEDIAN, PERCENTB, PSO, ROGERSSATCHELL, RVIR, SI, SQZMOM, STC, SWAK_2PHP, SWAK_BP, SWAK_BUTTER, SWAK_GAUSS, SWAK_HP, VIDYA, WAVETREND, ZIGZAG), the moving averages MA_Type.VIDYA and MA_Type.ALMA, and ten unstable-period ids. PERCENTILE's default period is now 100. - _ta_lib.pxd declares the new functions, ids and moving averages. - MA_Type, the unstable-period table, the function groups, _ta_lib.pyi and the docs list them. docs/funcs.md now lists every function: it also gains the 43 it was missing from earlier releases. The MINMAXINDEX stub returns int32. - TA_LIB_C_REQUIRED is 0.8.3, which the bindings now need. - _func.pxi, _stream.pxi, abstract.pyi and stream.pyi are regenerated. - Tests: the new unstable ids move their own function; under 'ALL' a function built on another unstable one (STC on EMA) grows by both; MA dispatches to VIDYA and ALMA; ZIGZAG's pivot index counts from the caller's first bar.
TA-Lib C 0.8.3's Auto-Stabilization takes an Auto level in place of a count:
talib.set_unstable_period('ALL', talib.UNSTABLE_AUTO_PREC_4) discards each
function's leading output until its first 4 significant digits no longer depend
on where the data starts (UNSTABLE_AUTO_PREC_8: 8). get_unstable_period()
returns the level. See https://ta-lib.org/api/unstable-period/#auto
…ses vForce
On Apple Silicon macOS, TA-Lib C 0.8.3 computes the batch call of the twelve
transcendental Math Transform functions (ACOS, ASIN, ATAN, COS, COSH, EXP, LN,
LOG10, SIN, SINH, TAN, TANH) with Apple's vForce, while a stream's update keeps
libm, so the two may differ in the last bits. test_stream.py failed 21 tests
there, which the macOS arm64 wheel build runs.
For those twelve, and only when TA_GetRuntimeInfo('vmath.transcendental')
reports a vector math library, the update-vs-batch comparison now allows equal
values or up to 8 ulps of the same sign, as TA-Lib C's own tests do. Every other
function and tier stays bit-exact. stream.py and the README say so.
Within a tolerance, so a change in how TA-Lib C computes them is caught by every wheel job. Refs TA-Lib/ta-lib#516
Without it TA-Lib C falls back to its scalar loop and every test still passes. Refs TA-Lib/ta-lib#516
TA-Lib C 0.8.3 is released. Source builds and the wheels now download and verify ta-lib-0.8.3-src.tar.gz, and the README's C install instructions and the build scripts' defaults name 0.8.3.
A real parameter now takes any real number, an int or a numpy integer or float included, and an integer parameter any integer, a numpy integer included, as talib.func does: abstract.BBANDS(d, 20, 2, 2) raised "expected float, got int". A value of a newly accepted kind is stored converted to float or int; a value accepted before is stored as given. A value of another kind, a float for an integer parameter included, still raises TypeError. Closes #430
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reviewing the dev branch with its changes.