Re: Add redis-db module to upstream?

[Bart Schaefer <schaefer@xxxxxxxxxxxxxxxx>]

On Jun 7, 11:29am, Oliver Kiddle wrote:
}
} Certrainly we want to take care that the hiredis dependency isn't making
} a zsh build more complicated. Perhaps the module could be disabled by
} default.

I believe this can be taken care of by configure, in the same way that
gdbm and pcre are left uncompiled if configure can't find them.

} I'd also agree with the points made by Bart such as unifying
} ztie, perhaps with a generic typeset option: typeset -o file=...

I think a separate builtin like ztie is the right way to go here,
rather than further overloading typeset.  We need a way to select
the module (ztie -d ...) and to specify external objects (ztie -f
and possibly other options if the notion of a "file" doesn't match
the usage).

} As for use, I've never actually got around to trying the gdbm module.
} What is it mostly good for? Very large variables or sharing variables?

Mostly it's for providing a shell-level interface to anything you would
store in a simple attibute-value database.  Specifically to zsh it can
be used to share state among shells (for example IF we could rely on it
gdbm woule be a much better file store for _store_cache/_retrieve_cache
than the current mechanism of serializing arrays).

If I understand redis's built-in clustering mechanism correctly, it
could be used to share state among shells on different hosts as well.