Fix an issue with bad allocation in post

[kilohsakul]

Fix of the problem with delta[q] reallocating post and screwing everything.

[Adda0]

I confirm that the tests run successfully on Linux.

[tfiedor]

Is this fix of #173 ?

[Adda0]

Yes, this is a fix for issue #173. However, this implements a different approach that what we have agreed on in our meeting. If we are OK with this approach instead, this PR would resolve #174, indeed.

[tfiedor]

Linked with #173 then

[tfiedor]

I have no major issues, I would strongly recommend to protect the static variable with const.

[tfiedor]

Is it worth keeping the code commented out?

[kilohsakul]

no, it was a working version, actually I don't know why the reviewers were assigned automatically

[Adda0]

I assigned them manually.

[kilohsakul]

Now it should be done, no more modifications planned.

[kilohsakul]

What should actually be the name of that function "mutable_post" ??
Something shorter? Like only delta.mutable(q)? delta.modifiable(q)? delta.modifier(q), ... or whatever else?

[Adda0]

I quite like mutable_post, actually. I am open to other suggestions, though.

We could work with something as mut_ref as “mutable reference”. nfa.delta.mut_ref(q) reads naturally to me and it warns us each time we use it that it is mutable and we have to be careful.

[Adda0]

If we are fine with this approach, the implementation seems OK to me.

[jurajsic]

It looks fine, if we are sure that the current usages of mutable_post cannot fuck up anything, then we can merge.

[Adda0]

if we are sure that the current usages of mutable_post cannot fuck up anything

In theory, it should not. This PR only makes it more obvious that the underlying structure can be modified. Therefore, one will not use the mutable accessor when they do not want to modify the data structure. Otherwise, using the mutable accessor behaves exactly as it did until now.

[kilohsakul]

Yes, this is a fix for issue #173. However, this implements a different approach that what we have agreed on in our meeting. If we are OK with this approach instead, this PR would resolve #174, indeed.

That's true, but this was really easy, and it looks ok. As I was saying at the meeting, I actually like it this more, but I thought it would be harder to do.

[Adda0]

this was really easy, and it looks ok.

I wholeheartedly agree. I am not against this approach. And as far as I know, I am not aware of any cases where this would not solve our problems.