I like to use cryptobib, because it gives me consistent results and it contains almost everything relevant to me.
However, as crypto.bib is over 725000 lines long, parsing it to create the bibliography takes a long time. This means that my $\LaTeX$ compile jobs take much too long.
I have now written a python script that will allow you to extract entries from large bibliographies.
This script will take your biblatex .