Loading ...
Sorry, an error occurred while loading the content.
 

Re: Does b:undo_ftplugin actually work?

Expand Messages
  • ZyX
    ... First, this is not the issue. Second, b: is *current* buffer scope. Not scope of the buffer being unloaded. -- You received this message from the vim_dev
    Message 1 of 23 , Sep 28 1:43 PM
      > Or to say it differently: by the time the BufUnload command is
      > triggered, the old buffer local variables have not been cleared yet, so
      > you can still access them.
      First, this is not the issue.
      Second, b: is *current* buffer scope. Not scope of the buffer being unloaded.

      --
      You received this message from the "vim_dev" maillist.
      Do not top-post! Type your reply below the text you are replying to.
      For more information, visit http://www.vim.org/maillist.php
    • Christian Brabandt
      Hi ZyX! ... Because those options aren t freed correctly. See the patch, I submitted. ... I guess we are not going to agree here. So just one final remark from
      Message 2 of 23 , Sep 28 1:53 PM
        Hi ZyX!

        On Fr, 28 Sep 2012, ZyX wrote:

        > суббота, 29 сентября 2012 г., 0:15:33 UTC+4 пользователь Christian Brabandt написал:
        > > >You may replace first "echom" with "setlocal" and see that
        > > > option is set for a wrong buffer. Your autocmd must not be used.
        > >
        > > You are changing buffers and wiping a buffer. I am not sure what you are
        > > trying to show here.
        > Do what I suggested with :setlocal. *It will set options for the buffer that **was not being unloaded***. ****Wrong**** buffer.

        Because those options aren't freed correctly. See the patch, I
        submitted.

        > And, by the way, you are using b:undo_ftplugin from the *wrong* buffer
        > as well. Because "b:" is *current* buffer scope and you must not touch
        > buffer that is not being unloaded.

        I guess we are not going to agree here. So just one final remark from me
        here, before I'll shut up.

        Those buffer-local variables will be freed afterwards. We have not yet
        fully unloaded the old buffer, therefore we can still undo the old
        options.

        regards,
        Christian

        --
        You received this message from the "vim_dev" maillist.
        Do not top-post! Type your reply below the text you are replying to.
        For more information, visit http://www.vim.org/maillist.php
      • ZyX
        ... To be more explicit: I do not want your autocommand to junk my ftplugin settings when I unload non-current buffer. And it happens every time I use bufkill
        Message 3 of 23 , Sep 28 1:56 PM
          суббота, 29 сентября 2012 г., 0:42:08 UTC+4 пользователь ZyX написал:
          > суббота, 29 сентября 2012 г., 0:15:33 UTC+4 пользователь Christian Brabandt написал:
          > > >You may replace first "echom" with "setlocal" and see that
          > > > option is set for a wrong buffer. Your autocmd must not be used.
          > >
          > > You are changing buffers and wiping a buffer. I am not sure what you are
          > > trying to show here.
          >
          > Do what I suggested with :setlocal. *It will set options for the buffer that **was not being unloaded***. ****Wrong**** buffer.
          >
          > And, by the way, you are using b:undo_ftplugin from the *wrong* buffer as well. Because "b:" is *current* buffer scope and you must not touch buffer that is not being unloaded.

          To be more explicit: I do not want your autocommand to junk my ftplugin settings when I unload non-current buffer. And it happens every time I use bufkill plugin, which I use almost always (just in case you thought it is uncommon) to keep windows intact (apparently, not only me; and not only this plugin uses this technique for the same purpose). And it will junk ftplugin settings whenever the current buffer has b:undo_ftplugin (i.e. has settings to junk), not the unloaded one (this increases amount of casualties).

          --
          You received this message from the "vim_dev" maillist.
          Do not top-post! Type your reply below the text you are replying to.
          For more information, visit http://www.vim.org/maillist.php
        • Gary Johnson
          ... I have autoindent set all the time, but it doesn t always work as I would like it to. For example, I may have a set of lines like these, each indented
          Message 4 of 23 , Sep 30 10:18 PM
            On 2012-09-28, ZyX wrote:
            > > Thanks for the reply, but I see that I didn't explain the problem
            > > very well. Also, some of my experiments created new buffers instead
            > > of replacing the contents of existing buffers with new filetypes, so
            > > I wasn't replicating the actual problem conditions.
            > >
            > > Let me try again.
            > >
            > > The actual problem is that I would like to set 'indentexpr' for
            > > buffers with no 'filetype' so that I get indenting behavior that I
            > > think might be useful when just opening Vim and typing notes. To
            > > that end I put the following in my ~/.vimrc.
            > >
            > > au BufWinEnter * if &ft == "" || &ft == "text"
            > > \ | setlocal indentexpr=indent(prevnonblank(v:lnum-1))
            > > \ | setlocal indentkeys-=o
            > > \ | let b:undo_ftplugin = "setl inde< indk<"
            > > \ | endif
            > You can just set 'autoindent'. This option is ignored in most
            > cases when using other ways to define indentation.

            I have 'autoindent' set all the time, but it doesn't always work as
            I would like it to. For example,

            I may have a set of lines like these,
            each indented from the one above,
            then I may add one ending in parentheses()
            then paste one below it line this.

            I would like to be able to type == to align that last line with the
            one above it, but == instead aligns that last line with the first
            line. The behavior has something to do with the parentheses.
            Without the parentheses, == aligns the last line with the one above
            it.

            I thought that my 'indentexpr' would be a good way to fix this, and
            it is except for my inability to unset it.

            I found a solution to that problem later in this thread.

            Regards,
            Gary

            --
            You received this message from the "vim_dev" maillist.
            Do not top-post! Type your reply below the text you are replying to.
            For more information, visit http://www.vim.org/maillist.php
          • Gary Johnson
            ... Here is the current solution, which has been working fine so far today. (I explained in a separate post how autoindent does not behave the same as this
            Message 5 of 23 , Oct 1, 2012
              On 2012-09-28, ZyX wrote:

              > > I'm just looking for a way to reset those local options when I edit
              > > a new file, b:undo_ftplugin seemed to be the way to do it, but it
              > > doesn't seem to do anything useful.
              > >
              > > Not using :edit is not a solution. For example, if I start vim
              > > and use ":MRU" to open a recently-used C file, I wind up with
              > > 'indentexpr' set as above, which wrongly indents C.
              > Unset 'indentoptions' on any filetype event, adding definition of
              > needed autocommand somewhere at the beginning of the vimrc (as it
              > is the first file sourced), definitely before “filetype plugin
              > indent on” or similar command, it should not interfere with
              > ftplugins then. Use 'autoindent' instead of your 'indentexpr', it
              > is here just for that reason.

              Here is the current solution, which has been working fine so far
              today. (I explained in a separate post how 'autoindent' does not
              behave the same as this 'indentexpr'.)

              au BufWinEnter * if &ft == "" || &ft == "text"
              \ | setlocal indentexpr=indent(prevnonblank(v:lnum-1))
              \ | setlocal indentkeys-=o
              \ | let b:undo_indent = "setlocal indentexpr< indentkeys<"
              \ | endif
              au FileType * setlocal indentexpr< indentkeys<

              Thanks for your help.

              Regards,
              Gary

              --
              You received this message from the "vim_dev" maillist.
              Do not top-post! Type your reply below the text you are replying to.
              For more information, visit http://www.vim.org/maillist.php
            • Bram Moolenaar
              ... I finally had time to look into this patch. It breaks the tests in a nasty way. Thus this is not the right solution. Instead of freeing the options they
              Message 6 of 23 , Nov 15, 2012
                Christian Brabandt wrote:

                > On Do, 27 Sep 2012, Gary Johnson wrote:
                >
                > > On 2012-09-27, ZyX wrote:
                > > > пятница, 28 сентября 2012 г., 8:04:05 UTC+4 пользователь Gary Johnson написал:
                > > > > I was working on some code that set b:undo_ftplugin, but it didn't
                > > > > have any effect when I set a new filetype. I copied
                > > > > $VIMRUNTIME/ftplugin.vim to ~/.vim and instrumented the section that
                > > > > is supposed to execute b:undo_ftplugin.
                > > > >
                > > > > func! s:LoadFTPlugin()
                > > > > echo "In s:LoadFTPlugin()" | sleep 2
                > > > > echo "b:undo_ftplugin is" b:undo_ftplugin | sleep 2
                > > > You should've put this line after the next one (or, better, remove
                > > > it as absense of the next message will indicate absence of
                > > > b:undo_ftplugin definition). And use ":echom", not ":echo |
                > > > sleep", then all messages will be seen when you do ":messages"
                > > > > if exists("b:undo_ftplugin")
                > > > > echo "b:undo_ftplugin exists" | sleep 2
                > > > > exe b:undo_ftplugin
                > > > > unlet! b:undo_ftplugin b:did_ftplugin
                > > > > endif
                > > > >
                > > > > Whenever I open a new file for which Vim can detect the filetype, or
                > > > > :edit some file, I always get the following errors.
                > > > b:undo_ftplugin is defined in filetype plugins. What else do you
                > > > expect? When opening new file and doing :edit buffer scope is
                > > > clean and filetype plugins are loaded by s:LoadFTPlugin function
                > > > *after* you test for b:undo_ftplugin. You should have used "set
                > > > filetype=new_filetype" on *existing* buffer to trigger desired
                > > > behavior.
                > > >
                > > > > Error detected while processing function <SNR>5_LoadFTPlugin:
                > > > > line 2:
                > > > > E121: Undefined variable: b:undo_ftplugin
                > > > > E15: Invalid expression: b:undo_ftplugin | sleep 2
                > > > >
                > > > > That is even after I execute
                > > > >
                > > > > :echo b:undo_ftplugin
                > > > >
                > > > > and verify that it is set correctly according to the new filetype.
                > > > Don't use :edit. It wipes the buffer.
                > > >
                > > > > It seems that b:undo_ftplugin does not exist in the environment in
                > > > > which s:LoadFTPlugin() is executed. Does setting b:undo_ftplugin as
                > > > > most current ftplugin scripts do actually do anything?
                > > > They do. Use "set filetype=new_filetype".
                > >
                > > Thanks for the reply, but I see that I didn't explain the problem
                > > very well. Also, some of my experiments created new buffers instead
                > > of replacing the contents of existing buffers with new filetypes, so
                > > I wasn't replicating the actual problem conditions.
                > >
                > > Let me try again.
                > >
                > > The actual problem is that I would like to set 'indentexpr' for
                > > buffers with no 'filetype' so that I get indenting behavior that I
                > > think might be useful when just opening Vim and typing notes. To
                > > that end I put the following in my ~/.vimrc.
                > >
                > > au BufWinEnter * if &ft == "" || &ft == "text"
                > > \ | setlocal indentexpr=indent(prevnonblank(v:lnum-1))
                > > \ | setlocal indentkeys-=o
                > > \ | let b:undo_ftplugin = "setl inde< indk<"
                > > \ | endif
                > >
                > > When I start vim and execute
                > >
                > > :echo b:undo_ftplugin
                > >
                > > I see
                > >
                > > setl inde< indk<
                > >
                > > as expected. Further, ":ls" shows
                > >
                > > 1 %a "[No Name]" line 1
                > >
                > > Now, if I open a C file, I expect to have 'indentexpr' empty, either
                > > because the C file is opened in a new buffer or because the C file
                > > was opened in the same buffer and the b:undo_ftplugin was executed.
                > > However, after executing
                > >
                > > :e foo.c
                > > :set indentexpr?
                > >
                > > I see
                > >
                > > indentexpr=indent(prevnonblank(v:lnum-1))
                > >
                > > and ":ls" shows
                > >
                > > 1 %a "foo.c" line 1
                > >
                > > I did take your advice about using echom and instrumented
                > > ftplugin.vim differently and verified that when it was executed by
                > > the 'filetype' change to "c", b:undo_ftplugin did not exist.
                > >
                > > If ":edit wipes the buffer" as you say, so that b:undo_ftplugin is
                > > deleted, then shouldn't that wiping reset the values of any local
                > > options?
                > >
                > > I'm just looking for a way to reset those local options when I edit
                > > a new file, b:undo_ftplugin seemed to be the way to do it, but it
                > > doesn't seem to do anything useful.
                > >
                > > Not using :edit is not a solution. For example, if I start vim
                > > and use ":MRU" to open a recently-used C file, I wind up with
                > > 'indentexpr' set as above, which wrongly indents C.
                >
                > Hi Gary,
                >
                > Looks like the initial buffer isn't correctly freed.
                > Try this patch: which at least also gets rid of the buffer local
                > options, when loading another buffer:
                > diff --git a/src/buffer.c b/src/buffer.c
                > --- a/src/buffer.c
                > +++ b/src/buffer.c
                > @@ -1702,6 +1702,12 @@
                > #endif
                > /* buf->b_nwindows = 0; why was this here? */
                > free_buffer_stuff(buf, FALSE); /* delete local variables et al. */
                > + /* can't set TRUE in free_buffer_stuff(), this would destroy the wininfo stuff,
                > + * so freeing the buffer options here afterwards manually */
                > + free_buf_options(buf, TRUE);
                > +#ifdef FEAT_SPELL
                > + ga_clear(&buf->b_s.b_langp);
                > +#endif
                > #ifdef FEAT_KEYMAP
                > /* need to reload lmaps and set b:keymap_name */
                > curbuf->b_kmap_state |= KEYMAP_INIT;
                >
                > Although, this doesn't solve the problem, that the b:undo_ftplugin
                > option isn't executed on BufUnload event. Possibly we need a global
                > BufUnLoad event, that takes care of this, e.g. something like this:
                >
                > " Make sure, the b:undo_ftplugin is also executed deleting the buffer
                > BufUnload * if exists("b:undo_ftplugin") | exe b:undo_ftplugin | endif
                >
                > But I am not sure, where to put this script. Perhaps also put this into
                > ftplugin.vim?

                I finally had time to look into this patch. It breaks the tests in a
                nasty way. Thus this is not the right solution. Instead of freeing the
                options they should be initialized to the global values.
                Perhaps setting buf->b_p_initialized to FALSE works?

                --
                The technology involved in making anything invisible is so infinitely
                complex that nine hundred and ninety-nine billion, nine hundred and
                ninety-nine million, nine hundred and ninety-nine thousand, nine hundred
                and ninety-nine times out of a trillion it is much simpler and more
                effective just to take the thing away and do without it.
                -- Douglas Adams, "The Hitchhiker's Guide to the Galaxy"

                /// Bram Moolenaar -- Bram@... -- http://www.Moolenaar.net \\\
                /// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
                \\\ an exciting new programming language -- http://www.Zimbu.org ///
                \\\ help me help AIDS victims -- http://ICCF-Holland.org ///

                --
                You received this message from the "vim_dev" maillist.
                Do not top-post! Type your reply below the text you are replying to.
                For more information, visit http://www.vim.org/maillist.php
              • Christian Brabandt
                Hi Bram! ... Unfortunately, that doesn t work. --git a/src/buffer.c b/src/buffer.c ... +++ b/src/buffer.c @@ -1703,6 +1703,7 @@ /* buf- b_nwindows = 0; why was
                Message 7 of 23 , Nov 16, 2012
                  Hi Bram!

                  On Do, 15 Nov 2012, Bram Moolenaar wrote:

                  > I finally had time to look into this patch. It breaks the tests in a
                  > nasty way. Thus this is not the right solution. Instead of freeing the
                  > options they should be initialized to the global values.
                  > Perhaps setting buf->b_p_initialized to FALSE works?

                  Unfortunately, that doesn't work.

                  --git a/src/buffer.c b/src/buffer.c
                  --- a/src/buffer.c
                  +++ b/src/buffer.c
                  @@ -1703,6 +1703,7 @@
                  /* buf->b_nwindows = 0; why was this here? */
                  free_buffer_stuff(buf, FALSE); /* delete local variables et al. */
                  buf->b_p_initialized = FALSE;
                  + buf_copy_options(buf, BCO_ENTER);
                  #ifdef FEAT_KEYMAP
                  /* need to reload lmaps and set b:keymap_name */
                  curbuf->b_kmap_state |= KEYMAP_INIT;

                  This also seems to work.


                  regards,
                  Christian
                  --
                  Besser schweigen und als Narr scheinen, als sprechen und jeden Zweifel
                  beseitigen.
                  -- Abraham Lincoln

                  --
                  You received this message from the "vim_dev" maillist.
                  Do not top-post! Type your reply below the text you are replying to.
                  For more information, visit http://www.vim.org/maillist.php
                • Bram Moolenaar
                  ... I ll try it out. Why BCO_ENTER and not BCO_ALWAYS, as it s used in the else part? -- If you put 7 of the most talented OSS developers in a room for a
                  Message 8 of 23 , Nov 17, 2012
                    Christian Brabandt wrote:

                    > On Do, 15 Nov 2012, Bram Moolenaar wrote:
                    >
                    > > I finally had time to look into this patch. It breaks the tests in a
                    > > nasty way. Thus this is not the right solution. Instead of freeing the
                    > > options they should be initialized to the global values.
                    > > Perhaps setting buf->b_p_initialized to FALSE works?
                    >
                    > Unfortunately, that doesn't work.
                    >
                    > --git a/src/buffer.c b/src/buffer.c
                    > --- a/src/buffer.c
                    > +++ b/src/buffer.c
                    > @@ -1703,6 +1703,7 @@
                    > /* buf->b_nwindows = 0; why was this here? */
                    > free_buffer_stuff(buf, FALSE); /* delete local variables et al. */
                    > buf->b_p_initialized = FALSE;
                    > + buf_copy_options(buf, BCO_ENTER);
                    > #ifdef FEAT_KEYMAP
                    > /* need to reload lmaps and set b:keymap_name */
                    > curbuf->b_kmap_state |= KEYMAP_INIT;
                    >
                    > This also seems to work.

                    I'll try it out. Why BCO_ENTER and not BCO_ALWAYS, as it's used in the
                    "else" part?

                    --
                    If you put 7 of the most talented OSS developers in a room for a week
                    and asked them to fix a bug in a spreadsheet program, in 1 week
                    you'd have 2 new mail readers and a text-based web browser.

                    /// Bram Moolenaar -- Bram@... -- http://www.Moolenaar.net \\\
                    /// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
                    \\\ an exciting new programming language -- http://www.Zimbu.org ///
                    \\\ help me help AIDS victims -- http://ICCF-Holland.org ///

                    --
                    You received this message from the "vim_dev" maillist.
                    Do not top-post! Type your reply below the text you are replying to.
                    For more information, visit http://www.vim.org/maillist.php
                  Your message has been successfully submitted and would be delivered to recipients shortly.