Re: [Emacs-orgmode] org-4.20: A few observations

[Carsten Dominik]

Hi Eric,

thanks for your report and careful observations.

> * Observations from using org-4.19d & 4.20
>
> ** Problem: table creating with bracket link with font-lock enabled

[...]

> 1) Create a table row with links: OK
>
> 2) with cursor at "x" press <TAB> to go to next row: PROBLEM
>
> 3) alignment problem persists as row are added
>
> 4) Moving cursor outside table and returning then typing <TAB> does
>    not resolve the problem.

> 5) However moving outside table and typing a char, killing line,
>    etc. then returning and typing <TAB> fixes alignment

First of all, if anything goes wrong with table realignment, you can 
always force a realign of the table with C-c C-c which in Org-mode 
generally has the meaning of "update" (exception: tags).

The org-mode table editor works by re-aligning the table each time you 
press TAB or RET in a table.  When the tables get large, this constant 
re-aligning does get kind of slow, so the table editor tries to be 
really smart and do the re-alignment only if necessary.  For example, 
it inserts and deletes characters in table fields, and if the column 
width is not affected, it does not trigger a re-align.  And it also 
tries to re-insert new rows without triggering a re-align.  This has 
worked really well so far, but indeed with hiding of link text the code 
inserting a new row without triggering re-align is broken.  For now I 
will fix it by triggering a re-align whenever a new line is inserted.  
When I get around to it I will fix the insertion code (which is a bit 
more complicated.

The reason why leaving the table, doing a buffer modification, getting 
back into the table and pressing TAB fixes the indentation is precisely 
because then Org-mode know the buffer has changed and a re-align might 
be necessary, so it triggers one.

> 6) Alignment after HTML export is correct ;)

Yes, this is fully independent.

> ** Problem: ispell hangs in org buffers with bracket links when font 
> lock is on
>
> N.B.  This is probably an old ispell.  Can someone running a
> newer version confirm this behavior?

I have tried quickly and not seen a problem with 3.6 - but I did not 
test exhaustively.

> ispell-version's value is
> "ispell.el 3.4 -- Fri Aug  4 09:41:50 PDT 2000"
>
> ** Problem: Strange edit behavior with links
>
> *** Kill-line behavior depends on font lock state with links in column 
> 0
> Consider the following pair of links with the cursor on the
> blank line between them, executing kill-line
> kills next two lines when font lock mode is on but only one line when
> font lock mode is off.
>
> [[http://www.google.com][Google]]
>
> [[http://www.google.com][Google]]
>
> *** Not a problem for links in column >1
>
>   [[http://www.google.com][Google]]
>
>   [[http://www.google.com][Google]]
>
>    [[http://www.google.com][Google]]
>
>    [[http://www.google.com][Google]]

This is interesting, but right now I don't know how to fix this.  Hmmm. 
 I guess this really is an Emacs bug.

> ** Observation: TODO sequences and dynamic setting of org-todo-keywords

When starting Org-mode, a number of regular expressions are computed 
and set.  Among others, there are the regular expressions matching the 
TODO keywords and other settings.  When you change the settings while 
Emacs is running, these expressions need to be re-computed.  M-x 
normal-mode does simply restart Org-mode and trigger this.  For the 
same money you could do M-x fundamental-mode and then M-x org-mode.  
The only way to avoid this would be to compute the regular expressions 
again before each command using them, which I find not acceptable.

The best way to experiment with TODO keywords in *not* putting Lisp 
code into the buffer and eval it with eval-region.   It is much better 
to use

#+TYP_TODO: TODO WAITING DONE

in the buffer and then press C-c C-c when you have changed one of those 
lines to update.

> If so, this may be something to put in the info files.  I am also
> aware that normal-mode is required after updating HTML style
> specifications.

I don't think this is true, is it?  Which specifications?

> However, it may be pleasing for newbies to understand the
> reason. C-h f normal-mode isn't particularly illuminating in this 
> regard.

I have tried to explain this well in the documents, but I guess it can 
be improved.  Yes I am not saying much about the "why", but you are the 
first to care :-)

I'll take another look at the documentation and see if this can be 
improved.

- Carsten