modifications to class_interface.c

[amoradinejad]

Hi @joezuntz

I noticed that you now replaced the old version of class_v2.7 with the latest version of class_v3.2. That's really great, I had to do it myself, but it's very nice that it is consistently replaced.

I have a few suggestions and a question about some modifications to class interface that may be helpful for some users, as I needed them myself.

1. Add the calculation of the matter transfer function for a given (k,z) to class_interface.py. To do this, one needs to make the following modifications:

• in boltzmann/class/class_v3.2.0/python/classy.pyxd add the definition of the new function of perturbations.c module in class_v3.2, perturbations_sources_at_k_and_z(), which calculates transfer function for a choice of (k,z)
• in boltzmann/class/class_v3.2.0/python/classy.pyx include a call to the above function, lets say called tk_lin() in analogy to pk_lin(),
• include a call to tk_lin() in boltzmann/class/class_interface.py.

2. Make the calculation of cmb optional. I found this important when running MCMC for galaxy clustering statistics only, since calculating the Cls makes the MCMC too slow at each step in the chain. So it would be nice in the output list of class_interface.py to have the option of only calculating matter power spectrum and transfer function, 'mPk mTk' as one can do when running class directly.
3. Would it be possible to have the extrapolation of the transfer function similar to what you have for the power spectrum? This would be helpful when calculating loops in galaxy statistics due to primordial non-Gaussianity, using FFTLog, which requires having kernels (which include the transfer function) at low and high k-values. In my version of cosmosis, I added a function to do this using the extrapolated p(k) and dividing it to primordial power spectrum and then building an interpolator for transfer function using init_interp_2d_akima_grid(), which extends the k-range covered by pk() extrapolator. But this seems too convoluted. There is perhaps a better way of doing this?

[joezuntz]

I haven't used Cython at all myself, so don't really know how to do your bullet point (1), but I'd be happy to accept contributions :-)

I can certainly do (2), that's straightforward. Then I can have a play with (3), after (1) is in place.

[amoradinejad]

Great, thanks for point 2.

As of point 1, I can send you the modified version of classy.pyxd classy.pyx. How's best to send them?

[joezuntz]

@amoradinejad if you know how to do a pull request then that's the best way to make sure your contribution is recorded. There's a "fork" button at the top right of the page, if you click that you'll get your own version on github. Then you can push your changes to that changed repository and then make a pull request from the tab with that name.

That's what I'd recommend, since it offer an easy way of reviewing the changes, as well as preserving your authorship of them. But you can also just email me the new versions of the file if you'd prefer not to.

[joezuntz]

I've added feature (2) - see the module.yaml file in the class directory for the option names.