convert array

[zonca]

hi, I would like to convert a long array from local frame to equatorial using: tpmu.convert

is there a better way than using a for loop? the best would be to have a function which accepts arrays for x, y and timetag.

thanks

[phn]

Hello,

I have re-written the C function convert, in convert.c, which is indirectly called by tpmu.convert and directly by tpm.convert, so that it accept coordinates and time tag as arrays. This should be faster than directly processing lists in python and then calling tpmu.convert for each coordinate pair in the list; haven't tested this claim though.

The changed code is in the git branch convert-input-an-array. I don't want to merge it into master at this time. I am currently writing documentation, and tests, and would like to make a release of the existing code with the documentation. After that, the changes to convert can be released as a new version.

Basically, we accept python lists into tpmu.convert and then construct C arrays from them, Then we pass the C arrays to tpm.convert. It returns C arrays, which are then processed back into python lists. These lists are then returned by tpmu.convert.

So now, we can do:

x_out, y_out = tpmu.convert(x=a_list, y=b_list, timetag=c_list, ..........) 
print tpm.fmt_alpha(utils.d2r(x_out[0]), tpm.fmt_alpha(tpmu.d2r(y_out[0])

Files test/sanity_test.py and test/test_convert.py have some simple examples.

Hope this helps.

Also, if you can send me a list of coordinates and time tags that you want to test, I can add that as a test and see what we get.

Thanks Prasanth

[zonca]

it looks wonderful, let me test it, then I'll reply back, thanks! andrea

[zonca]

tested, it works fine, however I expected a bigger improvement in performance, now on my laptop it takes 30s with lists of 46800 values. while with a python for loop using the master branch it takes 34.1s with the same lists.