Repository navigation
swt documentation mistake #838
Description
Documentation for pywt.swt states:
"If start_level = m is given, then the beginning m steps are skipped:
[(cAm+n, cDm+n), ..., (cAm+1, cDm+1), (cAm, cDm)]"
It is misleading, because level = n is implied to be positive integer and not non-negative integer, i.e. starts from 1 and not 0.
I provide a counterexample level=1, start_level = 1.
According to documentation, coeffs = [(cA1+1, cD1+1), (cA1, cD1)], len(coeffs) = 2.
In reality, coeffs = [(cA1, cD1)], len(coeffs) = 1.
I suggest removing (cAm, cDm) from documentation or better yet, changing indexes to start from 0 for clarity.
Right now default start_level = 0 corresponds to [(cAn, cDn), ..., (cA2, cD2), (cA**1**, cD**1**)] which is also convoluted and misleading.
Activity
rgommers commentedon Feb 20, 2026
Thanks for the report @Smogy001, you are correct that that doesn't look accurate.
This line is wrong:
From a quick look, I think it should read:
I think the subscript
1is used consistently throughout the docs, so this doesn't sound good to me.Smogy001 commentedon Feb 23, 2026
The line should read
[(cAm+n, cDm+n), ..., (cAm+1, cDm+1)]because minimumn=1.I agree that changing notation midway is undue. However the fact that start_level starts from 0 and level starts from 1 which corresponds to 2**0 scale bugs me.
rgommers commentedon Feb 23, 2026
Ah yes, you are correct - verified by running some code now. If you're interested in opening a PR to fix this, that'd be welcome.
Yes I understand that, but we can't really change it for backwards compatibility reasons. The intent of the code reads like the argument name should have been
skip_levelsrather thanstart_level.DOC: fix swt start_level coefficient range in docstring
DOC: fix the swt/swt2 `start_level` coefficient ranges