[Ifeffit] Small larch bug

[Matt Newville]

Hi Johan,



On Fri, May 27, 2016 at 6:21 AM, Johan Nilsson <johan.nilsson at chalmers.se>
wrote:

> Hello,
>
>
> I've found a small "bug" in larch when I tried to run the pre_edge
> function for some spectra. It seems that find_e0 will not accept a 2d
> 1-column array as input for energy or mu, and it gives a traceback which is
> quite difficult to decipher. Other functions like plot accepts these arrays
> without any problems, which made it a bit difficult to figure out what the
> problem was. Maybe there can be some some sort of warning if arrays have
> unexpected shapes, or an internal reshape if that works. I've attached a
> script that should give a minimum working example, my output is:
>
>
> larch> run('cu.lar')
> (612,)
> 8977.58
> (612, 1)
>   File "/usr/local/lib/python2.7/dist-packages/larch/__init__.py", line
> 37, in wrapper
>     return fcn(*args, **keywords)
>   File "/usr/share/larch/plugins/xafs/pre_edge.py", line 51, in find_e0
>     dmu = np.gradient(mu)/np.gradient(energy)
>   File "/usr/lib/python2.7/dist-packages/numpy/lib/function_base.py", line
> 918, in gradient
>     out[slice1] = (f[slice2] - f[slice3])file cu.lar, line 14
> IndexError: index 1 is out of bounds for axis 1 with size 1
>     find_e0(e,cu.mu,group=cu)
>   find_e0(e,cu.mu,group=cu)
> file cu.lar, line 14
> larch>
>

Thanks.  I'll turn this into a github Issue.  For completeness (though I
suspect you had figured this out), the immediate problem can be solved by
simply doing
     e = e.squeeze()

That said, I  do see your point that this should be done by find_e0().   If
every function that wanted a 1d array had to make such tests, it would
probably be a losing battle.  But, it is probably reasonable to expect the
"processing functions" like find_e0(), pre_edge(), and autobk() -- that is,
the functions designed to take "raw"-like data arrays" and turn them into
EXAFS -- should try to accept data arrays in lots of "formats", including
plain lists data with extra dimensions, and possibly even pandas dataframes.

--Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://millenia.cars.aps.anl.gov/pipermail/ifeffit/attachments/20160527/d122ad74/attachment.html>