pickling to strings?

[dimpase]

Sometimes it is better to have a pickle to a string, or to a memory area, e.g. in libgap setting, where GAP objects might be parts of a bigger structure that needs to be pickled, e.g. a Python one, cf https://docs.python.org/2/library/pickle.html

I imagine it should not be too hard to provide, as the existing in IO code surely has all the needed parts available, but is these a chance such a change is accepted and merged?

[olexandr-konovalov]

@dimpase I have IO_PickleToString and IO_UnpickleFromString (see https://gap-packages.github.io/scscp/doc/chap9.html#X84F055ED860120D5). Historically this appeared in SCSCP to satisfy its needs, and I am happy to submit a PR and move them to IO.

[dimpase]

This looks like a reasonable thing to do, thanks in advance!

[fingolfin]

PRs welcome :-)

[ChrisJefferson]

Stupid question, isn't what you want already there (this doesn't let you specify a memory block (edit: was group)). This has been there since 2011.

gap> s := IO_Pickle(Group((1,2,3)));
"PRMGILIS\>1PERM\>7(1,2,3)FAILFAIL"
gap> IO_Unpickle(s);
Group([ (1,2,3) ])

[ChrisJefferson]

Quickly checking, I notice this isn't documented, so that's a problem :)

[dimpase]

Document it, or it doesn't exist! :-)

And so we'd be busily picking GAP groups.

I don't understand what do you mean by saying that it doesn't let you specify a group.

[ChrisJefferson]

I meant existing memory block, which you mentioned, obviously had groups on the brain.

[ChrisJefferson]

Just in case you hit it, if you pickle permutation groups which are the results of GAP computations you might hit #91 ( which was just found by me, and fixed, a few days ago). This will only be a problem if you pickle the identity group, and only then the identity group returned by some GAP methods.