| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
| |
The two options are either using vis:message (the default) which will
keep all messages in the vis message buffer and include a header line to
distinguish the different messages.
The second option is using open and presenting messages in a separate
window supporting syntax highlighting.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
This is inline with our README.md claiming 'line' is the default
value for highlight_diagnostics.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |\
| |
| |
| |
| | |
restore the primary cursor position after a workspaceEdit
See merge request muhq/vis-lspc!47
|
| | | |
|
| | | |
|
| | | |
|
| |/ |
|
| |\
| |
| |
| |
| | |
Fix lscp.open_file not actually adding to history
See merge request muhq/vis-lspc!45
|
| |/ |
|
| |
|
|
|
|
|
|
|
| |
Using the new latex lexer from the upcoming vis release I noticed
that vis-lspc default style id 43 overrides the environment style
from the lexer.
To prevent id clashing use the highest allowed id for lexer styles,
which is 64 according to the enum defined in vis/ui.h.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The last commit introduces support for highlighting empty selections
by properly calculating their vis ranges.
However File:content does return an empty string for empty ranges
and we can not check whether the diagnostic still applies to the same
position.
This is fixed by faking a one character range when receiving an
empty one.
Doing so we get a valid one-character content we can compare and
still have proper highlighting because we unconditionally decrement
the range end in highlight_diagnostics.
Fixes: 7099e56062861fc8dbb143e78447cce6bf886a97
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Some language servers report diagnostics with empty ranges, start and
end are at the same line and character, which currently results in
no highlighting at all.
Since the end of a LSP range is exclusive we always subtract 1 from
it which results in invalid ranges when the server has sent an empty one.
Introduce a check to fix this special case.
`textDocument/publishDiagnostics` doesn't always have a `code` field,
resulting in `diagnostic.code` being nil and so it is replaced with
`diagnostic` in those cases.
|
| | |
|
| |
|
|
|
| |
Since the diagnostics table is created when we open the file
it will always be ~= nil.
|
| | |
|
| |
|
|
|
|
|
| |
* fix typos
* wrap file or project names in code blocks
* drop the hint taht vis-lspc is in an early state
* s/json/JSON/
|
| |
|
|
|
|
| |
This is implemented by the two new commands:
* lspc-next-diagnostics - jump to the next available diagnostic
* lspc-prev-diagnostics - jump to the previous available diagnostic
|
| |
|
|
|
|
|
|
| |
Range highlighting used the vis_range and line highlighting the LSP
range. Do not touch old code you no longer understand because you think
you are apparently smarter than your past self!
Fixes: efd5262f2c45d3289e482b1091d3c1d44cf67a42
|
| |\
| |
| |
| |
| |
| |
| | |
use new vis 0.9 features
Closes #43
See merge request muhq/vis-lspc!42
|
| | |
| |
| |
| |
| |
| | |
Default formatting options should not be configured by the LSPC client,
the user should either configure them explicitly or use something
like editorconfig or mode lines.
|
| | |
| |
| |
| | |
Since vis 0.9 the options are exposed via the vis.options table.
|
| |/ |
|
| | |
|
| | |
|
| |\
| |
| |
| |
| |
| |
| | |
try to convert workspaceEdit.documentChanges to plain changes
Closes #34
See merge request muhq/vis-lspc!44
|
| |/
|
|
|
|
|
| |
We do not support workspaceEdit.documentChanges but some LSP servers
do not respect our capabilities and send them anyways.
Try to optimistically convert simple documentChanges to plain changes.
We only support simple TextDocumentEdit[].
|
| |
|
|
|
|
|
|
|
|
|
| |
The method textDocument/formatting is allowed to return null/nil which
was directly passed to vis_apply_textEdits causing an exception
when trying to iterate the textEdits table.
Only pass the result of textDocument/formatting to vis_apply_textEdits
if there is an actual result.
Fixes: e83950a7149f364b653395cdd4d9446875a7f469
|
| |
|
|
|
| |
If the result is MarkedString[] the table must have a length > 0,
because all returned object types are represented as lua tables.
|
| | |
|
| | |
|
| | |
|
| |\
| |
| |
| |
| | |
Check modified
See merge request muhq/vis-lspc!41
|
| | | |
|
| | | |
|
| |\|
| |
| |
| |
| | |
Add textDocument/formatting implementation
See merge request muhq/vis-lspc!40
|
| | |
| |
| |
| |
| |
| |
| |
| | |
This should be wired up to code that's in a pull request in the vis
repository for the default options. This PR does not yet expose
`expandtab`.
Referenced PR: https://github.com/martanne/vis/pull/717
|
| | |
| |
| |
| |
| |
| |
| | |
This is tested with the Rust language server and the Lua one. Moving
the cursor to the correct position afterwards works with Rust, but not
with Lua. Being smarter about the ranges is probably necessary for Lua
and might make other ones more robust.
|
| |/ |
|
| | |
|
| | |
|
| |
|
|
|
| |
When using vis_open_doc_pos, don't reopen the file if the path of the
opened file and the doc_pos match.
|
