For windows users .... click here doesn't do anything

This includes making sure it's read in properly in Funcs.py as well as in the main program.

Need to add a vanilla FoF thing so users who don't want to trust the graph theory can make use of the traditional HG82 version of the FoF algorithm without some features

• no updating position by averaging
• no multiple runs (not needed since removing the position averaging will ensure all runs are identical)

Really we only need to parallelize the "FoF" function over the number of runs which are taking place. Maybe this can even be a user option? Multiprocessing = True or something like that.

I don't know what the maximum we could do but we should just do that.

This really only effects users who want to run the algorithm many times and average over using graph theory, which may not be everyone but the Vanilla FoF algorithm should be included for people who want to use just that.

So lets parallelize the FoF function in the Py2Friends.py That should bring 100 runs down to 10 minutes or so, which it is now in anycase. Plus the Graph Theory is already pretty fast after being vectorized properly and should be able to average over all the runs in about 30 - 90 seconds.

There are two instances of <> in the Py2Friends.py program that cause a syntax error immediately.

My guesses are
if len(checklimit)<>0: should be if len(checklimit)>0:
while list(friends_after) <> list(friends_before) and iterations<20: should be while list(friends_after) > list(friends_before) and iterations<20:.

Need to flag problematic cases where large groups are found as a single group because of a small number of connections between the 2.

We have done this previously by post processing the output files but this should be part of the main program such that it is independent of output.

We need documentation for users

So far for 45 000 galaxies the algorithm takes about 10-15 minutes to run 10 times. Ideally at least 100 runs would be done meaning about 100 minutes.

Need to speed up the overall time of the actually FoF algorithm.

Obvious solution is to parallelize running the different trials but I think we can speed up the whole thing before we parallelize and then future proof using parallelization.

For 45 000 galaxies, the time taken should be under 2 mins for 100 runs. This would mean future large surveys with 100's of thousands of galaxies would be handled well.

I don't want to use C :'(

I think the best way to get the param file implemented into the package is to create a param file generator that will spawn a param file template which the user can then update.

The param file can be generated blankly or alternatively we can include some default values. Specifically for specific type of surveys with Shecter params included.

The main body of the program, running the algorithm is actually very short. Writing the numerous outputs takes up the most space by far. Perhaps a class with .totxt() .tocsv() .toIAPUC() .toPartiView() etc. Anytime someone makes a request for a new type of output type we should be able to just add it to the class?

For now those three should do well.