render-complete: if at EOF, there won't be a string, so be careful.

Signed-off-by: NeilBrown <neil@brown.name>

Add notifier to non-home panes on a doc.

Use this to clean up and install a new document when the old one dies.

Signed-off-by: NeilBrown <neil@brown.name>

doc_destroy: don't close the pane before sending the destroy message to it.

Signed-off-by: NeilBrown <neil@brown.name>

doc-dir: locate the editor before destroying the pane which lets us find it.

Signed-off-by: NeilBrown <neil@brown.name>

Make sure to initialize all 'cmd_info'.

Missed this one.

Signed-off-by: NeilBrown <neil@brown.name>

Change doc panes to have an allocated 'data'.

This allows us to store other stuff than just the document.

Signed-off-by: NeilBrown <neil@brown.name>

discard doc_first_mark in favour of vmark_first.

Signed-off-by: NeilBrown <neil@brown.name>

Change popup to store the doc rather than the pane for the temp document.

This makes it easier to eventually destroy.

Signed-off-by: NeilBrown <neil@brown.name>

change doc_new to return a doc.

some callers want to set the name, and that currently requires the actual doc.

Signed-off-by: NeilBrown <neil@brown.name>

Discard doc_get_pane() in favour of doc_from_pane()

The pane was (almost) never used other than to extract the document,
so just return the document.

This will allow us to change the ->data stored on a document pane.

Signed-off-by: NeilBrown <neil@brown.name>

Add 'void *misc' field to cmd_info

Use this for passing the 'struct stat' for doc:same-file.

Signed-off-by: NeilBrown <neil@brown.name>

Get rid of rl_mark and lots of casts.

Now that all marks have mdata, we don't need this extension.

Signed-off-by: NeilBrown <neil@brown.name>

Get rid of 'struct point'.

Just use 'struct mark' which now as a void* mdata for extension.

Signed-off-by: NeilBrown <neil@brown.name>

Move size field from point to point->links.

Now a point is nearly the same as a mark.

Signed-off-by: NeilBrown <neil@brown.name>

Remove ->doc field from point.

No longer needed.

Signed-off-by: NeilBrown <neil@brown.name>

use doc_get_pane to find document in search functions.

..because I don't want point to hold the doc anymore.

Signed-off-by: NeilBrown <neil@brown.name>

in doc-dir we can use ->home->data to get document.

And use ->mark to find point too.

Signed-off-by: NeilBrown <neil@brown.name>

doc_attr: pass pane instead of doc

Signed-off-by: NeilBrown <neil@brown.name>

Point_notify_change: pass doc explicitly, don't use point.

Signed-off-by: NeilBrown <neil@brown.name>

Remove use of point->doc from point fore/back to mark.

No longer needed.

Signed-off-by: NeilBrown <neil@brown.name>

Remove doc arg from doc_next_mark.

It is no longer needed.

Signed-off-by: NeilBrown <neil@brown.name>

tlist_for_each{,_entry}_continue* no longer need head of list.

We can use the type information to stop at the right place.

Signed-off-by: NeilBrown <neil@brown.name>

Remove unused function point_reset

Signed-off-by: NeilBrown <neil@brown.name>

simplify/improve mark-at-point functions.

Signed-off-by: NeilBrown <neil@brown.name>

Pass pane to doc_set_attr instead of using pt->doc;

Signed-off-by: NeilBrown <neil@brown.name>

Remove ->owner field from point.

No longer needed.

Signed-off-by: NeilBrown <neil@brown.name>

Points: allocate list of links separately.

It was just too much hassle having a single allocation and managing
a single pointer to a point.

Signed-off-by: NeilBrown <neil@brown.name>

Remove ci->pointp

Now completely gone.

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointless doc and pointp from emacs_file_complete and emac_doc_complete

Signed-off-by: NeilBrown <neil@brown.name>

emacs_findfile: use pane_attr_get instead of doc_attr.

Signed-off-by: NeilBrown <neil@brown.name>

emacs_kill_doc - no pointp

Signed-off-by: NeilBrown <neil@brown.name>

Remove last pointp from render-lines

Signed-off-by: NeilBrown <neil@brown.name>

Stop using pointp for Move-View-Pos

Signed-off-by: NeilBrown <neil@brown.name>

complete_return doesn't need pointp either.

Signed-off-by: NeilBrown <neil@brown.name>

Change complete_set_prefix to no need direct access to point/doc

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp usage from emacs insert/delete commands.

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp usage from a lot of Move-* calls.

Signed-off-by: NeilBrown <neil@brown.name>

Remove doc arg from mark_to_mark.

It isn't really neded.
We can easily remove doc from doc_next_mark_all() as it isn't used.
We can also have a version of doc_prev_mark_all() that only gets
called when we *know* there is a previos mark, as we do ni mark_to_mark.

Signed-off-by: NeilBrown <neil@brown.name>

Send point as ci->mark for events from keyboard or mouse.

I am deprecating ->pointp, but sometimes I need that mark.

Signed-off-by: NeilBrown <neil@brown.name>

Don't set pointp for "Replace"

no replace function uses it anymore.

Signed-off-by: NeilBrown <neil@brown.name>

Change "Refresh" function to pass 'point' via ci->mark.

Signed-off-by: NeilBrown <neil@brown.name>

Change render-line command to accept point in 'mark2' instead of 'ptp'.

And get rid of point pointp stuff.

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp arg for render-line-prev

We only needed the doc, and we can get that from the pane.

Signed-off-by: NeilBrown <neil@brown.name>

Stop using pointp in text_reundo and text_replace.

use home->data and home->point instead.

Signed-off-by: NeilBrown <neil@brown.name>

text_reundo: rename 'm' to 'm2'.

I want to introduce a more global 'm'.

Signed-off-by: NeilBrown <neil@brown.name>

Move final doc default actions from 'view' to 'doc'.

Signed-off-by: NeilBrown <neil@brown.name>

Improve default move-file

Instead of moving one char at a time, just jump to the right end.

Signed-off-by: NeilBrown <neil@brown.name>

Now that move_X is in core-doc, remove the pointp references.

Signed-off-by: NeilBrown <neil@brown.name>

move simple word/char/eol/file movement from 'view' to 'doc'.

I think it makes more sense here.

Signed-off-by: NeilBrown <neil@brown.name>

remove two pointp users from emacs-search.

Just use doc_replace to trigger a refresh.

Signed-off-by: NeilBrown <neil@brown.name>

Remove render-dir and render-text

Increasingly dead-weight.  Now that we have render-lines and
render-format, these can go.

Signed-off-by: NeilBrown <neil@brown.name>

use doc_get_pane to get document for render-format movement.

Removes another pointp user

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp from render_hex_eol.

Signed-off-by: NeilBrown <neil@brown.name>

Use panes and marks to communicate with CountLines

so no more pointp;

Signed-off-by: NeilBrown <neil@brown.name>

Do for dir_close what we did for docs_bury.

It really is the same thing.

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp setting from mark_same_pane

completely pointless

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp usage from text_save_file and text_get_str

Signed-off-by: NeilBrown <neil@brown.name>

Notfiy:Replace doesn't need pointp any more.

So discard it.

Signed-off-by: NeilBrown <neil@brown.name>

Use "Notify:Replace" for notifications.

We currently use "Notify" both to make changes and to notify
of changes.
This can be confusion - they are different tasks needing different names.

Signed-off-by: NeilBrown <neil@brown.name>

docs_destroy - avoid pointp.

Use doc_get_pane instead.

Maybe I should be sending a message....

Signed-off-by: NeilBrown <neil@brown.name>

Change doc:destroy to work with pane rather than point.

Signed-off-by: NeilBrown <neil@brown.name>

Remove pane_point.

Only one called remains, so open-code it there.

That caller should eventually go too.

Signed-off-by: NeilBrown <neil@brown.name>

Remove dependence of search on pane_point.

This requires PointDup being able to provide a simple mark,
and a new function to find the document pane.

Signed-off-by: NeilBrown <neil@brown.name>

"Close" no longer needs pane_point.

render_lines was the last users, and it now uses panes for vmarks.

Signed-off-by: NeilBrown <neil@brown.name>

Change vmark functions to take a pane instead of a document.

Signed-off-by: NeilBrown <neil@brown.name>

Change vmark_matching to take a pane rather than a doc.

Signed-off-by: NeilBrown <neil@brown.name>

Render-attach functions no-longer need the pane_point.

So discard it.

Signed-off-by: NeilBrown <neil@brown.name>

Change doc_{add,del,find}_view into commands.

they get send to the document instead of having to find it.

Signed-off-by: NeilBrown <neil@brown.name>

Register keymaps early rather than late.

The more we use them, the more careful we must be.

Signed-off-by: NeilBrown <neil@brown.name>

Change mark_same to key off tile rather than point.

Thus remove pointp reference from render-lines-notify

Signed-off-by: NeilBrown <neil@brown.name>

Have all "render-line" function fail cleanly if point/mark are NULL.

This makes it a bit cleaner to find the render-line pane with a void call.

Signed-off-by: NeilBrown <neil@brown.name>

Provide a "Move-to" command to use e.g. when completing a search.

Better than direct access to point_to_mark().

Signed-off-by: NeilBrown <neil@brown.name>

Change default_render from a struct field to an attribute.

Signed-off-by: NeilBrown <neil@brown.name>

Use Move-File to move to start of a newly created document.

This is cleaner than moving the point directly

Signed-off-by: NeilBrown <neil@brown.name>

text: fit bug with text deletion.

I wasn't quite leaving things is a good state.
I think this is better - as this ref no longer over-takes
a follow ref that won't be relocated.
But I really need to review this code.

Signed-off-by: NeilBrown <neil@brown.name>

emacs: use pane_final_child rather than pane_point.

This is (now) a suitable place to attach attributes.

Signed-off-by: NeilBrown <neil@brown.name>

Introduce doc:set-name

Use this variously in mode-emacs to avoid pane_point calls.

Signed-off-by: NeilBrown <neil@brown.name>

Remove pointp from render_format_attach

Not used.

Signed-off-by: NeilBrown <neil@brown.name>

Use the new "ThisPane" to find pane to load new doc/file.

Makes the code nicer too.

Signed-off-by: NeilBrown <neil@brown.name>

Introduce ThisPane

This is similar to "OtherPane" but find the current tile.

popup uses this if it want to make a popup that is over a doc instead of
over the display.

This removes another use of ->point

Signed-off-by: NeilBrown <neil@brown.name>

Introduce PointDup command

This make it possible to get a dup of point without
explicitly finding the point. Makes it more indirect.

You this in 'search' to avoid hunting up the tree.

Signed-off-by: NeilBrown <neil@brown.name>

Don't make new point directly in doc_attach

sometimes we want point_new sometimes point_dup

Signed-off-by: NeilBrown <neil@brown.name>

Split a 'doc' pane aways from the 'view' panes.

They really are different things.

Signed-off-by: NeilBrown <neil@brown.name>

use new doc_attach_view for attaching "view"

This is a common task and it is good to centralize it.

Signed-off-by: NeilBrown <neil@brown.name>

Split doc_attach_view from doc_open.

It is silly combining these two functions with very different results.

Signed-off-by: NeilBrown <neil@brown.name>

use pane and mark to pass document to doc:set-attr

Signed-off-by: NeilBrown <neil@brown.name>

Change doc:get-attr to use a pane rather than a point.

Signed-off-by: NeilBrown <neil@brown.name>

Use doc_notify_change rather than point_notify_change were possible.

This gives more flexibility going forward.

Make sure doc_notify_change is just as fast when a point is provided.

Signed-off-by: NeilBrown <neil@brown.name>

Use pane rather then point to pass doc to doc:step

Signed-off-by: NeilBrown <neil@brown.name>

Use pane instead of point to pass doc to "Release"

Signed-off-by: NeilBrown <neil@brown.name>

Change doc:set-ref to take a pane rather than a point.

Signed-off-by: NeilBrown <neil@brown.name>

Change doc_find to return a pane.

That fits the pattern better.

Signed-off-by: NeilBrown <neil@brown.name>

Remove the document list and use a list of panes instead.

The 'null display' (ed->root.focus) is now the parent of multiple
panes, each of which owns a document (->data) with a backlink
(doc->home->data == doc).

Now in many places we pass around this pane instead of a point when we
just want to identify a document and not a location in the document.
This means we don't need to allocate points so much, or fake up
temporary ones.

doc_attach() must be used at document creation to attach a document to
the tree.

So
 - "*Documents*" watches this list of siblings rather than the original list.
 - commands can be sent to these panes using key_handle_focus.  Some
   doc commands can be sent to any pane.
 - doc creation returns the pane of the document in ci->focus
   rather than a fake point
     e.g. same_file, load_file

 - pane_attach() takes a pane, not a point.
 - doc_open no-longer needs a gross hack
 - editor_choose_doc returns a pane, and so doesn't need a point.
 - popup holds the doc pane, not a point.

etc.

Signed-off-by: NeilBrown <neil@brown.name>

Change 'editor' to be the root of the pane tree.

Displays are all children of that root.

Signed-off-by: NeilBrown <neil@brown.name>

Change doc:reundo and doc:get-str to use the pane-path.

Signed-off-by: NeilBrown <neil@brown.name>

send "doc:replace" up the pane path

Reduces need for passing points and documents around.
There will be more of this.

Signed-off-by: NeilBrown <neil@brown.name>

text: fix bug with refresh when text added to start of first line.

The return value didn't match the comment...

Signed-off-by: NeilBrown <neil@brown.name>

text: fix undo/redo bug.

we move the point ref forward, and them move along the mark list
to match.
But we need to go *beyond* any mark at the same position
else when we step the ref forward we won't know to keep going....
It is confusing, but this doesn't crash.

Signed-off-by: NeilBrown <neil@brown.name>

Remove the doc_operations structure complete.

After all, it is empty now.

Signed-off-by: NeilBrown <neil@brown.name>

Make 'step' a command.

This is the most-used command, I really have to optimise this one day.

Signed-off-by: NeilBrown <neil@brown.name>