When :NERDTreeFind creates a tree, it resolves the file's path, and
makes that its root. See :h resolve() for more info; it basically
tracks down links to their ultimate source. Then :NERDTreeFind
tries to find the file under that root, so it can be revealed. The
problem is that it compares the file's unresolved path against the
resolved path in the root. Here is the scenario:
/tmp/ is a link to /private/tmp/
:e /tmp/foo will open the file as expected
:NERDTreeFind will first create a tree with a root of /private/tmp/
Then it tries to find /tmp/foo, but it can't, because the path
names don't match.
This commit resolves /tmp/foo to the actual location of
/private/tmp/foo; thus, it is able to find the file in the tree.
* Force sort to recalculate the cached sortKey.
The problem in issue #880 was caused by the sort using the old sortKey.
For example, given nodes A, B, and C, if B were renamed to D, the sort
was still using B as its sortKey, thus not moving it.
It's a bit of a hack, but if we set g:NERDTreeOldSortOrder to an empty
list, the cached sortKey will be recalculated. I did the same thing for
both the Copy and Add functions as well.
* Add a comment to explain the let ... = [] statement.
When middle-clicking, the s:TreeFileNode.GetSelected() function is
called along the way. If this is done outside of the NERDTree window,
there is not "b:NERDTree" variable, and the "E121: Undefined variable"
exception is thrown. This function was trying to catch only the NERDTree
specific errors; thus, it let the Undefined variable exception slip by.
This commit causes the function to catch all errors.
* Use a better arg for FINDSTR when using the m,l command in Windows.
* Remove trailing spaces.
* Use vim getf...() functions in place of DOS commands.
* Remove unnecessary variable.
* Put labels in the text of the file/dir listing.
This PR is a rewrite of the "s:UI.getLineNum()" function. This change
improves the logic and clarity of this function, and it fixes the function's
ability to work with cascades.
Fixes#529.
This change puts the burden on the user to make sure
g:NERDTreeBookmarksFile is correctly set and the path it contains is
present and has the proper permissions. If this is not the case, an
error message is displayed. This is a better solution than to blindly
create the path, when it may actually be impossible to do so.
When the user has `'autochdir'` turned on, opening a new NERDTree will
cause the current working directory to change. To prevent this
happening, remember the CWD and reset it if NERDTreeFocus caused it to
change.
When NERTDtree opens a file, it compares this new file to all open
buffers to see if it's already open. If the user has 'ignorecase' turned
on then the comparison of "file" and "File" says they're the same, and
NERDTree won't reopen the file. This commit forces a case sensitive
comparison by using the ==# operator.
If the cursor is not already positioned on a node when the bookmark
table is closed, then the resulting position of the cursor is not
determined. Here, we default to positioning the cursor on the
root when the bookmark table is closed.
Also add a new global variable to track when the g:NERDTreeSortOrder
changes. If it has been changed, or when the cached _sortKey value is
uninitialized, then calculate the sort key. This improves processing
speed over the previous commit, and allows on-the-fly changes to the
sort order, (without required vim to be restarted.)
The call to AddDefaultGroupToSortOrder in NERD_tree.vim is redundant
because it's also done every time sortChildren is called. And since the
check is done only once, there's no need for a function either.
sortChildren now just contains the needed if statement.
This commit makes adding, removing and finding key maps an O(1)
operation instead of O(n), where n is the number of pre-existing maps.
In my testing, averaged over 100 iterations, this reduces the time spent
initializing NERDTree at Vim startup from ~73ms to ~9.7ms. That's with
only the default included key maps.
