Sort diagnostics according to severity for detailed diag responses

[bstaletic]

Users are more likely to be interested in errors than warnings in /detailed_diagnostic responses. That means we need to sort the diagnostics before finding the one nearest to the cursor.

For LSP completers, that's easy as the severity is numerical. For Omnisharp-roslyn and TSServer, the severity is a word and we can't just say sort( key = severity ).

Since LSP severity was convenient for comparison, I have opted for sorting C# and TS diags by the equivalent LSP severity.

---

This change is 

[codecov]

Codecov Report

Merging #1728 (5fb0927) into master (0ce532e) will increase coverage by 0.40%.
The diff coverage is 66.66%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1728      +/-   ##
==========================================
+ Coverage   95.05%   95.46%   +0.40%     
==========================================
  Files          51       83      +32     
  Lines        6711     8194    +1483     
  Branches        0      163     +163     
==========================================
+ Hits         6379     7822    +1443     
+ Misses        332      322      -10     
- Partials        0       50      +50     

[bstaletic]

According to the coverage data, we're only ever dealing with errors in our tests, at least for C# and TS.

[puremourning]

Change LGTM, but could we at least add a regression test for LSP ? Seems like something that we could easily forget about.

Reviewable status: 0 of 2 LGTMs obtained

[puremourning]

Reviewed 3 of 3 files at r1, all commit messages.
Reviewable status: 0 of 2 LGTMs obtained (waiting on @bstaletic)

[bstaletic]

I wanted to, but I couldn't make a test case without pulling in the entirety of ranges-v3. That just feels like a massive dependency for our need.

Reviewable status: 0 of 2 LGTMs obtained (waiting on @bstaletic)

[puremourning]

Don't we just need some code which produces a warning and error on the same line.

I think clangd does this. maybe something like:

int x;
if ( x = 2 )
{
}

With -Wno-error -Werror=uninitialized

[bstaletic]

Tests added.

Reviewable status: 0 of 2 LGTMs obtained (waiting on @puremourning)

[puremourning]

Reviewed 6 of 6 files at r2, all commit messages.
Reviewable status: 0 of 2 LGTMs obtained (waiting on @bstaletic)

---

ycmd/completers/language_server/language_server_completer.py line 1666 at r2 (raw file):

    current_column = lsp.CodepointsToUTF16CodeUnits(
        GetFileLines( request_data, current_file )[ current_line_lsp ],
        request_data[ 'column_codepoint' ] - 1 )

I'm not sure about this....

the help doc for this func says the offset is in 1-based units:

def CodepointsToUTF16CodeUnits( line_value, codepoint_offset ):
  """Return the 1-based UTF-16 code unit offset equivalent to the 1-based
  unicode codepoint offset |codepoint_offset| in the Unicode string
  |line_value|"""

If current_column is 0-based then we should subtract 1 from the result of the call, not the input.

[bstaletic]

Reviewable status: 0 of 2 LGTMs obtained (waiting on @puremourning)

---

ycmd/completers/language_server/language_server_completer.py line 1666 at r2 (raw file):

Previously, puremourning (Ben Jackson) wrote…

I'm not sure about this....

the help doc for this func says the offset is in 1-based units:

def CodepointsToUTF16CodeUnits( line_value, codepoint_offset ):
  """Return the 1-based UTF-16 code unit offset equivalent to the 1-based
  unicode codepoint offset |codepoint_offset| in the Unicode string
  |line_value|"""

If current_column is 0-based then we should subtract 1 from the result of the call, not the input.

Without the -1, current_column is 1-based.
However, we are coparing that column to diagnostic[ 'range' ], later in the for loop, where diagnostics have 0-based character offsets. The comment I added above the call to CodepointsToUTF16CodeUnits() should explain that.

I remember going back and forward and not being sure if that -1 should go before of after ) of the function call.

Thanks for catching this! Fixed.

[puremourning]

Reviewed 1 of 1 files at r3, all commit messages.
Reviewable status: 1 of 2 LGTMs obtained (waiting on @bstaletic)

[mergify]

Thanks for sending a PR!