Added support for rangeFormatting

[ryuukk]

Fixes Pure-D/code-d#241

[WebFreak001]

thanks for the addition. This is quite a complex change and also changes how the previous full formatting works.

Because of this, this PR should really add some unittests for the new diff method.

[WebFreak001]

add compare functions / operators for Position so this is easier to read

[ryuukk]

I'm not sure how i should write this either xD

I think it's best to keep this simple, it's not needed elsewhere

[WebFreak001]

in the struct Position add:

    int opCmp(const Position other) const
    {
        if (line < other.line)
            return -1;
        if (line > other.line)
            return 1;
        if (character < other.character)
            return -1;
        if (character > other.character)
            return 1;
        return 0;
    }

this allows you to compare positions like a < b

I believe it would be simpler to just have

editRange.start <= formatRange.end && editRange.end >= formatRange.start

here.

[ryuukk]

I can make the "formatting" untouched, and only "rangeFormatting" uses the new diff method, so it doesn't change the all behavior at all

Also i contacted DLS author about taking some of their ideas, he agreed about shipping the code from GPL3 to MIT

LaurentTreguier/azure-dlang-test#1

[WebFreak001]

I think it's ok for the full formatting method to use diff as well (it improves bandwidth), just needs some tests

[ryuukk]

I'm not sure how i should write the tests

[WebFreak001]

all structs are plain old data. You can simply add a unittest block after the method and call it with your custom test data and run the tests using dub test.

[ryuukk]

I reverted the old behavior, because i don't know exactly what to do with tests, i don't want to mess up things that were working previously, i'm not familiar enough with the project

If someone wants to change the old behavior, it could be done on a separate PR

[WebFreak001]

ok I have added a few unittests, the diff algorithm really does work amazingly for how simple it is! I'm squashing it and merging it in directly on master + my changes afterwards, so I'll close this once done. (with reference to merge commit)

[WebFreak001]

merged in with 95c14e9

[WebFreak001]

here are some commits you may be interested in checking out what I changed in your code:

• 303e261 - added tests in this commit
• 2a65f69 - fix format diff for utf32 characters
• f830dca - reduce complexity of format diff for large files

you initially used offsetToPosition (which is UTF-16 codepoints) instead of bytesToPosition, which caused bugs with non-ascii code, so I changed that and then optimized using the movePositionBytes which reuses calculation of previous positions, which is essential for big files.

It would be possible to actually include the position calculation inside the diff code, as then it wouldn't need to iterate over the whole file twice, but I think it's ok how it is now.