Rick Frankel <rick@rickster.com> writes:

> Eric-
>
> I have debugged why the :colnames argument is not respected as to
> removing the header from a table var.
>
> * Given
> #+name: with-hline
> | A | B | C |
> |---+---+---|
> | 1 | 2 | 3 |
> | 4 | 5 | 6 |
>
> #+name: with-hline2
> | B | C | D |
> |---+---+---|
> | 2 | 3 | 4 |
> | 5 | 6 | 7 |
>
> #+name: emacs-echo
> #+BEGIN_SRC emacs-lisp :var table=with-hline :colnames yes
> (mapcar (lambda (x) (mapcar '1+ x)) table)
> #+END_SRC
>
> #+call: emacs-echo(table=with-hline2)[:colnames yes]
>
> * Evaluation
> When the `#+call' line is executed, `org-babel-process-params' is
> called multiple times. When it is called the last time to process
> the actual input (here, calling =emacs-echo= with the param
> =with-hline2=), the `:colname-names' parameter has already been (to
> =(table "A" "B" "C")=, taken from the variable in the source blocks
> header) as a side-effect of the call to
> `org-babel-get-src-block-info'. So, when the block is executed later
> in the function, then next call to `org-babel-process-params' does
> not call disassemble table (taking into account the value of
> `:colnames') but uses =processed-vars= directly instead, causing the 
> full
> table to be used without calling `org-babel-disassemble-tables'.
>
> Here's the relevant section of code from `org-babel-process-params':
>
> #+BEGIN_SRC emacs-lisp
> (if (and (assoc :colname-names params)
> (assoc :rowname-names params))
> (list processed-vars)
> (org-babel-disassemble-tables
> processed-vars
> (cdr (assoc :hlines params))
> (cdr (assoc :colnames params))
> (cdr (assoc :rownames params))))
> #+END_SRC
>
> And here's a cleaned-up debug trace:
>
> #+BEGIN_EXAMPLE
> (list processed-vars)
> ;; called again later in execute-src-block, does not properly parse 
> =with-hline2= var
> org-babel-process-params(((:comments . "") (:shebang . "") (:cache . 
> "no") (:padline . "") (:noweb . "no") (:tangle . "no") (:exports . 
> "results") (:results . "silent") (:var . "table=with-hline2") (:hlines . 
> "yes") (:padnewline . "yes") (:session . "none") (:colnames . "yes") 
> (:result-type . value) (:result-params "replace") (:rowname-names) 
> (:colname-names (table "A" "B" "C"))))
> ;; calls disassemble-tables and returns with :colname-names set
> org-babel-process-params(((:comments . "") (:shebang . "") (:cache . 
> "no") (:padline . "") (:noweb . "no") (:tangle . "no") (:exports . 
> "results") (:results . "replace") (:var . "table=with-hline") (:colnames 
> . "yes") (:session . "none") (:padnewline . "yes") (:hlines . "yes")))
> org-babel-get-src-block-info()
> org-babel-execute-src-block(nil nil ((:var . "table=with-hline2") 
> (:results . "silent")))
> org-babel-ref-resolve("emacs-echo(table=with-hline2)[:colnames yes]")
> org-babel-ref-parse("results=emacs-echo(table=with-hline2)[:colnames 
> yes]")
> org-babel-process-params(((:comments . "") (:shebang . "") (:cache . 
> "no") (:padline . "") (:noweb . "no") (:tangle . "no") (:exports . 
> "results") (:results . "replace") (:var . 
> "results=emacs-echo(table=with-hline2)[:colnames yes]") (:colnames . 
> "no") (:hlines . "yes") (:padnewline . "yes") (:session . "none")))
> org-babel-lob-execute(("emacs-echo(table=with-hline2)[:colnames yes]" 
> nil 0 nil))
> #+END_EXAMPLE
>
> rick

Hi Rick,

Thanks for taking the time to find the root of this problem.  I believe
I've fixed this by change the `org-babel-merge-params' function so that
when the value of a variable is update, then colname-names and
rowname-names values saved for that variable are removed.

In my local tests the attached patch fixes this issue.  If it works for
you as well then I'll apply it.

Thanks,