Hi,

Thanks again!

Nicolas Goaziou <mail@nicolasgoaziou.fr> writes:

>> I did not know markers but they seem perfect in this case.  The manual
>> mentions setting markers to nil after use.  I guess it's not necessary
>> here since they are in a (let ⋯)?
>
> It is. Binding between the symbol and the marker disappears with the
> `let', but the marker still exists. It cannot be GC'ed unless it is set
> to nil.
>
> It is not terribly important here as we're working in a temporary buffer
> anyway, so the marker will ultimately disappear at the end of the export
> process. However, it's a good habit to have.

Thanks for the explanation.

>> I check if I'm at a footnote reference 'cause I never want to deal with a
>> footnote-*definition* directly.  AFAIK org-footnote-re matches both.  It
>> seems a footnote-reference can never be at the beginning of line, but I
>> would still prefer a more explicit test through org-element.
>
> I wasn't clear. The check is important. The minor issue is that you bind
> LABEL and FOOTNOTE-TYPE too early, before making sure you are at
> a footnote reference. It would be more logical to do
>
>   (let ((object (org-element-context)))
>     (when (eq (org-element-type object) 'footnote-reference)
>       (let ((footnote-type (org-element-property :type object))
>             (label (org-element-property :label object)))
>         ...)))

I see.  I disagree that it's more since it's directly inside a loop over
org-footnote-re.  So if we are not at a footnote-{reference,definition}
it's probably a bug in the regexp. 

>> The only "bug" *I'm aware of* is that it will pick up the wrong new-label
>> for footnote for something like [fn:ref with space].  But this is anyway
>> not a proper label, I think.  Is that OK?
>
> [fn: ref with space]

While your comment excels in preciseness the terseness makes it hard to
appreciate its depth.  In my org-installation "[fn: ref with space]" is
not a valid footnote.

>> +	      (when (and (not included) (> (hash-table-count footnotes) 0))
>> +		(org-with-wide-buffer
>> +		 (goto-char (point-max))
>> +		 (unless (bolp) (insert "\n"))
>> +		 (maphash (lambda (ref def) (insert (format "[%s] %s\n" ref def)))
>> +			  footnotes)))))))))))
>
>   (unless included ...)
>
> is sufficient. No need to check for table emptiness (although it doesn't
> cost much): maphash will simply do nothing.

Thanks.  It's not necessary anymore.

>> +      (let ((marker-min (point-min-marker))
>
> This marker is not necessary since you're not going to add contents
> before (point-min) anyway. A plain number is enough.

I might if I include *Bar here:

* Foo
[1] foo

* Bar
Baz[1]



>> +	(goto-char (point-min))
>> +	(while (re-search-forward org-footnote-re nil t)
>> +	  (let* ((reference (org-element-context))
>> +		 (label (org-element-property :label reference))
>> +		 (digit-label (and label (org-string-match-p "\\`[0-9]+\\'" label))))
>> +	    (when (eq (org-element-type reference) 'footnote-reference)
>
> See above about order of bindings.

Adopted and commented on above.

>> +	      (goto-char (1+ (org-element-property :begin reference)))
>> +	      (when label
>> +		(let ((new-label
>> +		       (buffer-substring-no-properties
>> +			(point)
>> +			(progn (if digit-label (insert (format "fn:%d-" id))
>> +				 (forward-char 3)
>> +				 (insert (format "%d-" id)))
>> +			       (1- (search-forward "]"))))))
>
>> +		  (unless (eq (org-element-property :type reference) 'inline)
>
> A comment about what we're going to do would be nice.

Comments are added.

—Rasmus

-- 
I feel emotional landscapes they puzzle me