Contribution

[Almesi]

Hey there, like i said yesterday i would like to contribute. I started some time ago with my own Standard library. https://github.com/Almesi/AL_StdLib
If you find anything in there that is worth your interest i would gladly help you. I also got some other ideas which may be interesting.
Do you have a policy of what you would like to include in your Library and what not?

[sancarn]

Hi Almesi,

I've added Contribution.md which should help with some generic coding style guidance.

As for things I'm looking for:

• Data structures like Stack, Heap, BinaryTree, QuadTree, R-tree, Graph, Digraph will always be beneficial.
  • Additionally algorithms on top of these like Graph: A* search, Min spanning tree, shortest path, max flow etc.
• Linear Algebra stuff like Matrix, Vector, Cross product, Dot product, ... may be of benefit. Would maybe be worth looking at using OpenSL for some of these.

There is also a roadmap which you can look on to see if there are any specifics which interest you :)

[sihlfall]

Hi sancarn,
thank you for all your work on this great project. I see you are looking for a native Regex implementation. I do have one, which I could contribute. It's rather unpolished yet, but I could bring it to a good shape, if you like.
With best regards,
sihlfall

[sihlfall]

What do you mean by generating data? Can you give an example?

[sancarn]

Sure, so this is something I started doing in stdRegex2. E.G.

With stdRegex.Create("[A-Z]-\d{5}")
  Debug.Print .Generate()  'F-46215
  Debug.Print .Generate()  'K-95642
  ...
End With

Mainly useful for creating dummy datasets. The code I was using for this is moderately simple but not 100% sure how easily this could be applied to your own engine.

[sihlfall]

I'd say (wild guess)

• it's simple for regexs without asserts (^, $, \b etc.) or lookarounds. You just traverse the abstract syntax tree left-to-right,
• it's a bit more tricky for regexs with asserts;
• it's possible, but a huge amount of work for regexs with lookarounds, yet without backreferences: You can transform the regex with lookarounds into a regex without lookarounds by intersecting NFAs. I remember that was discussed somewhere on Stackoverflow.
• I'd say it's nearly impossible for regexs with both lookarounds and backreferences, as in general, regexs with backreferences cannot be represented as an NFA, so we cannot get rid of the lookarounds. On the other hand, I've nowhere seen any theoretical result saying that lookarounds make regexs with backreferences more powerful, hence my statement is really only a wild guess.

The more practical problem I see is that you have to specify some probability distribution if there is more than one alternative. E.g., with what probabilities should a, aa, aaa, aaaa etc. be generated if the regex is a+? And should . really generate every possible Unicode character?

[sancarn]

@sihlfall Is there any way that I can help push this to a PR? I've been a little out of the loop. Looking at https://github.com/sihlfall/vba-regex/ everything looks ready? It just needs compiling and a PR submitting? Or is that here ready to go? https://github.com/sihlfall/stdVBA/blob/regex-vbaonly/src/WIP/stdRegex3/stdRegex3.cls

If so happy for you to promote this to /src and push a PR :) It'd be useful if we can make a github action to compile and automate any future PRs too :)

[sihlfall]

Dear sancarn,
sorry for not answering earlier, I was totally under water with other projects.
I try to keep stdRegex3.cls in sync with the master branch of vba-regex, so either one should be fine to use. The dev branch of vba-regex is a bit ahead (in particular, it supports unbounded modifiers). I am planning to do a merge shortly, after running all the tests.
Feel free to go with any one of these, I am pretty confident that the project is in a good state.
Best regards,
sihlfall