Evolve the rule for directory in ignore_files to be more natural
and easy to understand.
When a final '/' or '/$' is found we consider the pattern to match
a directory and the pattern is not modifed. In turns, is used, before
matching a directory's name a final '/' is appended to its name
before checking if it matches the pattern.
With the previous rule a final '/' in the pattern meant also a directory
but the '/' was removed from the pattern.
When a user's or project's module configuration file is changed we
make sure that the config.plugins fields are all restored so that
all plugins already loaded can continue to work.
Some asserts are placed in case that can effectively occur
so we remove the assertion and we return false. In turn we
adapt the logic accordingly so when false is returned to add
a watch we do not open that directory.
Works correctly and the logic seems sound even if somewhat quirky.
`^%.` match any file of directory whose basename begins with a dot.
`^/node_modules$/"` match a directory named `node_modules` at the project's root.
Note that the final '/' needs to be at the end. The '/' after the '^' needs to be there to trigger
a match of the full path filename so we are sure it is at the root.
PROBLEM: the '/' to trigger full path match could be in a pattern's special expression like:
[^/]
`^%.git$/` match any directory name '.git' anywhere in the project.
`^/%.git$/` match a directory named '.git' only at the project's root.
`^/subprojects/.+/` match any directory in a top-level folder named "subprojects".
`^/build/` match any top level directory whose name begins with "build"
PROBLEM: may be surprising, one may expects it matches only a directory named 'build'. It actually acts like
it was `^/build.*/`.
When changing or opening a project directory do not
take the selected item from suggestion but simply the
entered text as it is.
Otherwise the user may be unable to choose a directory
if the text matches the beginning of suggestion.
Close#791
For special file types like the ones in /dev/ the info
entry's type is neither file neither dir.
We prevent these kind of files from being listed in the
project.
Attempt to fix issue #791.
The logic set with the previous commit for suggest_directory
is similar to the one we use except the previous expression
was false do to operator precedence for "and" versus "or".
With the modification here, when opening a project directory,
we suggest the recently used projects
if the text is equal to dirname(project_dir) + "/" which
happens to be the text the command view is initially set to.
In addition we do the same if text is "". If the condition is
not met we return the suggestions from common.dir_path_suggest to
match the text entered.
Works well on Linux but may not solve the problem on Windows, it
should be tested.
Avoid reloading the core.keymap module when user's config
or project module change.
The reason is the plugins like autocomplete can add keymaps
and the additions from plugins would be lost.
Close issue #793
There was a double error because the config.ignore_files was
used at two differect places in different ways.
Now we apply coherently the original rule to apply
config.ignore_files to the basename of each file or directory.
It wasn't fine to call core.open_doc without filename argument
and later call Doc:save without providing both the filename and
the absolute filename. It was giving a Doc in an inconsistent
status where self.filename was set but not self.abs_filename.
Added an asset to detect early the problem if ever happens again.
In turn the problem prevented the project's module hook to work if the
file was newly created.
Simplifies and uniformize the logic on the Lua side for the
setting of directories' watches. Now we always use the methods:
systems.watch_dir_add / rm
on all the project's directories at any depth when we are not
in files limit mode.
In files limited mode the functions systems.watch_dir_add/rm are
called only on the expanded folders. The shown_subdir table is also
updated only in files limited mode.
On the C side, using the dmon library, we remove the recursive argument
from the system.watch_dir and we always call it recursively except on
Linux. At the same time the functions:
systems.watch_dir_add / rm
are provided but as dummy functions that does nothing except on Linux
where they work as before to add / remove sub-directories in the inotify
watch.
In this was on the Lua side we always act we if the watches needed to be
set for each sub-directory explicitly, independently of the system.
The important improvement introduced is that we always avoid calling
dmon_watch recursively on Linux. This latter thing is problematic with
inotify and is therefore avoided on Linux.
On the other side we simplifies the logic on the Lua side and remove
conditions based on the OS used.
The NagView takes some actual space in the Y and when it appears
it cause the documents' content to be displaced.
The movement of the documents' content is annoying and should be
avoided so we draw the NagView entirely in overlay mode using defer
draw and we always keep its y size to zero to don't affect the
other application contents.
Changes in project's module required an application restart to work.
Now the project will be re-scanned when the project's module changes.
In addition ensure borderless window config is changed when changed
in user's preferences.
It is not a good practice to keep a reference to the project's
directory object outside of the "core" module itself.
The TreeView was using such a reference in the cache item for each
file or directory entry. Replace the reference to the object with
the absolute name of the project directory.
Address issue:
https://github.com/lite-xl/lite-xl/issues/689
Attempt to provide a more accurate fix to commit:
59f64088e1
For this latter what happens is that any change inside a directory
cause the corresponding entry to be folded in the TreeView.
The new change is more accurate because we remove only the stale
entry corresponding to the delete event and we do not reset the
cache of the parent directory using the modify event.
Fix a conspicuous omission to call the dmon_unwatch function
when changing project directory.
This uncovered a bug or a quirk of the dmon library where the watch_ids
can change as a result of calling dmon_unwatch because they are just
indexes on a contiguous array. Use a workaround to always unwatch the
first valid watch_id N times.
Address issue:
https://github.com/lite-xl/lite-xl/issues/689
Attempt to provide a more accurate fix to commit:
59f64088e1
For this latter what happens is that any change inside a directory
cause the corresponding entry to be folded in the TreeView.
The new change is more accurate because we remove only the stale
entry corresponding to the delete event and we do not reset the
cache of the parent directory using the modify event.
The function "file_search" in core.init was sometimes giving a wrong index
value, off by one.
The problem happened for example when the entry to search was "less than"
the first entry, the function returned a value of two instead of one as
expected.
The bug was easily observed creating a new directory with a name that comes
as the first in alphabetical order within the project.
The function "file_search" in core.init was sometimes giving a wrong index
value, off by one.
The problem happened for example when the entry to search was "less than"
the first entry, the function returned a value of two instead of one as
expected.
The bug was easily observed creating a new directory with a name that comes
as the first in alphabetical order within the project.
When `highlighter:insert_notify` was called, a hole in the array was
created.
If another call to `highlighter:insert_notify` happened before the hole
was filled, a `Position out of bounds` error could have been raised.
When adding a directory in a project we check if the filesystem is too
slow. If it is too slow we act as if the projects was files-limited by
the number of files but we show a specific warning.
This solution is not perfect but for very low filesystem it can limit
the problem. Otherwise the application would be totally irresponsive.
Fix a problem introduced when fixing the dirty pixel problem, commit
cb08c5c. The node, when determining the layout was rounding the size
of the fixed-size view. In turns this latter was calling move_towards
to the default_size it wanted. If default_size was non-integer the
value vas never archieved because it was rounded during layout and
move_towars was keeping the editor busy by setting the
core.need_redraw flag.
On windows paths belonging to network volumes will be gives like:
\\address\share-name\path
Now the code recognize these paths and treat them correctly.
Fixing the Node's clipping rectangle make the clipping in DocView:draw()
partially redundant. This latter is now no longer needed to clip
on the right when drawing the document's lines but it still serves to
the purpose of clipping on the left, before the gutter region.
The last column of pixel on the window's right side isn't correctly
drawn and pixels appear dirty and more noticeably when the a NagView
message was previously shown, a stripe of red pixels remains on the right.
We use now a more souding roundig scheme. Now the rectangles to clip or to
draw are passed around as Lua numbers without any rounding. In turns, when
the rect coordinates are passed to the renderer we ensure the border of the
rect are correctly snapped to the pixel's grid. It works by computing the
coordinates of the edges, round them to integers and then compute the rect's
width based on the rounded coordinates values.
The syntax highlighter keep a cache of the documents like tokenization.
In order to minimize the amount of tokenize re-computations we insert some
emtty lines or remove some lines in the highlither lines corresponding to
the lines added or removed to the document.
In the highlither thread We should accept a previously generated line tokenization
past first_invalid_line only if the text is the same. The text can change because of
insert or remove operations.
Close#573.
The syntax highlighter keep a cache of the documents like tokenization.
In order to minimize the amount of tokenize re-computations we insert some
emtty lines or remove some lines in the highlither lines corresponding to
the lines added or removed to the document.
In the highlither thread We should accept a previously generated line tokenization
past first_invalid_line only if the text is the same. The text can change because of
insert or remove operations.
Close#573.
Bring back the command like before to keep single selection but with
ctrl+f3 keybindings. Change the name of the new multi-cursor command
but keep the ctrl+d keybinding.
The initial position for the search is defined by the last selection
towards the end of the file.
After reaching the end of the file, it would always select the same
selection to start the search from.
Now, we start the search from each selection, until a new occurrence is
found.
When using hidden suggestions remember the text user was typing when
navigating suggestions.
Ensure also that in the previously searched expressiosn we have no
duplicate entries.
If the selected text containes newlines it doesn't make sense to
use it as the initial text in the "replace text" command view.
Do not use the selected text if a newline is found in the selection.
Fix#511.
This function gets called at every `core.step`, so we should avoid
having to recalculate the scrollable size every time, as it could get
very expensive on long lines.
There are really multiple things here in the close_all_docviews
function:
1. we reset the Node's tab_offset
2. we ensure the core's active_view is properly set
3. we close LogViews as well as DocViews
Some conditions seems to never happen but we stay safe and try
to cover all possible cases.
Francesco Abbate
Adam
Guldoman
takase1121
Jan200101
Jipok
Joshua Minor
Rongfei Wang
Timofffee