Re: [PATCH] lisp/org-table.el: Allow named columns on lhs

[Gavin Downard <gavin.downard@runbox.com>]

Max Nikulin <manikulin@gmail.com> writes:
> On 19/07/2023 09:36, Gavin Downard wrote:
>> +++ b/lisp/org-table.el
>> @@ -2253,8 +2253,7 @@ LOCATION is a buffer position, consider the formulas there."
>>   			((not (match-end 2)) m)
>>   			;; Is it a column reference?
>>   			((string-match-p "\\`\\$\\([0-9]+\\|[<>]+\\)\\'" m) m)
>> -			;; Since named columns are not possible in
>> -			;; LHS, assume this is a named field.
>> +			;; This is either a named field or column.
>>   			(t (match-string 2 string)))))
>>   		    (rhs (match-string 3 string)))
>>   		(push (cons lhs rhs) eq-alist)
>
> Notice
> 		   "Double definition `%s=' in TBLFM line, please fix by hand"
>
> below. A bit more code is required to keep this sanity check for named columns.
>

Oh, good catch. Specifically, I think this should be caught inside of
`org-table-recalculate', where it catches conflicting direct column
references (eg "$1") and end-relative column references ("$<").
> 	    let* ((rhs (org-table-formula-substitute-names
> 			 (org-table-formula-handle-first/last-rc (cdr eq))))
> 		   (old-lhs (car eq))
> 		   (lhs
> 		    (org-table-formula-handle-first/last-rc
> 		     (cond
> 		      ((string-match "\\`@-?I+" old-lhs)
> 		       (user-error "Can't assign to hline relative reference"))
> 		      ((string-match "\\`\\$[<>]" old-lhs)
> 		       (let ((new (org-table-formula-handle-first/last-rc
> 				   old-lhs)))
> 			 (when (assoc new eqlist)
> 			   (user-error "\"%s=\" formula tries to overwrite \
> existing formula for column %s"
^ right here
> 				       old-lhs
> 				       new))
> 			 new))
> 		      (t old-lhs)))))

Also, this should probably be addressed in a different patch, but the
above code doesn't catch two different end-relative column references
that refer to the same column, such as "$<" and "$>>" in a two-column table.

I have also modified the manual to reflect the addition of named
columns; I'll send an updated patch after I add the check for double definitions.

Max Nikulin <manikulin@gmail.com> writes:
>
> `if-let' is not available in Emacs-26
>

Are you sure? It looks like `if-let' was introduced in Emacs 25.