Back to the main page.

Bug 2957 - revamp functionality to convert between different representation of freq-domain data

Reported 2015-09-07 21:16:00 +0200
Modified 2018-11-12 10:58:12 +0100
Product: FieldTrip
Component: core
Version: unspecified
Hardware: PC
Operating System: Mac OS
Importance: P5 normal
Assigned to: Jan-Mathijs Schoffelen
Depends on:
See also:

Jan-Mathijs Schoffelen - 2015-09-07 21:16:21 +0200

Inspired by Craig, I looked a bit into the state of the code that deals with the conversion of different representations of frequency domain data. This occurs mainly in ft_checkdata, private/univariate2bivariate (which relies heavily on ft_checkdata), and prepare_freq_matrices. The code is pretty outdated (e.g. still relying on old style ft_selectdata, and making some hardcoded assumptions / requiring an explicit datatype etc.) and deserves to be updated. This bug is now filed as a placeholder to keep track of the changes. At this point in time I see the following potential improvements: - prepare_freq_matrices should be rewritten to rely on ft_selectdata and ft_checkdata, rather than using its own code. JM actually has a rewritten version lying around somewhere (from several years back), which may serve as a template. There's actually a separate bug assigned to JM that is about this topic. - univariate2bivariate needs a massive clean up. Since JM probably is the only one who understands this function, he should do it. - ft_checkdata (subfunction fixcsd) can have some slightly extended functionality to support an efficiently conversion between 'fourier' and 'sparse', without the need to keep the individual observations in memory in an intermediate step (i.e. equivalent to the 'fullfast' option (this is a feature request by Craig).

Jan-Mathijs Schoffelen - 2016-02-21 14:43:02 +0100

intermediate update: point 1, the revamp of prepare_freq_matrices is implemented. The rest is still open.

Jan-Mathijs Schoffelen - 2018-11-12 10:57:56 +0100

no priority here it seems, closing for now.