[PATCH v3 0/8] emacs: JSON-based search cleanups

[Mark Walters]

On Sun, 15 Jul 2012, Austin Clements <amdragon at MIT.EDU> wrote:
> On Sun, 15 Jul 2012, Mark Walters <markwalters1009 at gmail.com> wrote:
>> On Sun, 15 Jul 2012, Austin Clements <amdragon at MIT.EDU> wrote:
>>> This version swaps out the notmuch-search-do-results macro for a
>>> higher-order function, notmuch-search-foreach-result.  This requires
>>> less squiting to understand and clearly distinguishes the arguments
>>> passed in to the function from the arguments passed to the callback.
>>> This version also updates the docstring for
>>> notmuch-search-result-format to mention that multi-line result formats
>>> work and how to enter them, and it adds a NEWS patch.
>>
>> Hello
>>
>> I am afraid I have found a minor (but reproducible) bug in the line
>> re-drawing even with single line results. Find a search result with a
>> partially elided author field and put the cursor after the ellipsis in
>> that line. Then update the tags. The result gets redrawn with ellipsis
>> written out in full. Re-redrawing with the cursor after the author field
>> redraws the line with the keeping the ellipsis written out in full,
>> whereas re-redrawing the line with cursor before the author field gets
>> it written with the correct ellipsis.
>
> Arrrg, overlays.
>
> I can think of two ways to fix this.  We could generate the author
> elision overlays lazily (say, via jit-lock).  This would have the added
> benefit of eliminating what I think is the last quadratic factor in
> building search buffers, but there are much easier ways to do that.  Or,
> I could scrap the insert-before-markers nonsense and manipulate point
> directly in notmuch-search-update-result, with the caveat that the
> little bit of support it had for doing sane things in some situations
> involving save-excursions would be lost.  Given that we never call
> notmuch-search-update-result inside a save-excursion (precisely because
> I couldn't reliably hit the narrow window of situations it could handle
> when there were save-excursions involved), I'd lean toward the latter
> option.

I think I don't really know enough emacs/lisp to be able to say anything
sensible. I think manipulating point directly seems good because then I
think we can make sure things work in the multiline case too. 

I haven't yet worked out whether multiline is an amusing "oh look it works"
or genuinely useful thing. I am leaning towards the latter as do often
work on laptops with quite small screens and think I would like to see
more about a smaller number of results.

Best wishes

Mark