Determine if binary data files are still necessary

[ids1024]

Words uses binary data files generated from plain text input for performance. This adds complexity that may be unnecessary on modern hardware.

[ids1024]

@mk270 What do you think of this?

[mk270]

What's the benefit of changing this? Is the complexity making it harder to package the system?

Open to changing this eventually, and/or providing alternatives, but we need to be really careful.

[ids1024]

The specific reason I thought of this is because it would help for my Android app. Since the binary format is platform specific (if I'm not mistaken), supporting multiple platforms would require multiple copies of the data. Consequently my app only supports arm currently, which is generally sufficient for Android, but having x86 as well would be good with Google bringing Android apps to Chromebooks.

It would also simplify the build process, but that is a fairly small thing.

[mk270]

I think you're right - there are lots of little reasons for making this change; I'd like to get the test suite sorted before changing the input format, for obvious reasons.

If we use SQLite, we'll be creating a new dependency, with possible licensing implications(*). We could do this as a separate binary, which you could ship instead of words. This is what I've done for the ZeroMQ/JSON service, too.

(*) I have a vague recollection that whereas SQLite is pseudo-public-domain, the Ada bindings for it are GPL. BICBW.

[darkestkhan]

What licensing implications? sqlite is public domain.

[mk270]

@darkestkhan there are Ada SQLite bindings in GNATcoll, which is licensed under GPLv3. Are you aware of other SQLite bindings for Ada? Am I misunderstanding something, e.g., what "bindings" are?