Re: [BUG] org-plot: Unable to use text xtics with type:2d (+ more) [9.7-pre (N/A @ /home/viz/lib/emacs/straight/build/org/)]

[Visuwesh <visuweshm@gmail.com>]

[வெள்ளி மே 24, 2024] Ihor Radchenko wrote:

> Visuwesh <visuweshm@gmail.com> writes:
>
>> Unless I misunderstood the code, the line
>>
>>       ;; Check type of ind column (timestamp? text?)
>>       (when (plist-get params :check-ind-type)
>>
>> should be
>>
>>       ;; Check type of ind column (timestamp? text?)
>>       (when (plist-get (cdr type) :check-ind-type)
>>
>> because (1) org-plot/collect-options only adds a select number of
>> keywords to the plist and :check-ind-type is not a part of the select
>> members, and (2) org-plot/gnuplot is never called with a non-nil value
>> for the optional argument PARAMS in tree.
>
> I do not think that it is right.
> AFAIU, the idea is that `org-plot/preset-plot-types' provides some
> default options, but the user can overwrite these defaults in the #+PLOT
> line. What you propose will disregard the values of
>
>  :set :line :map :title :file :ind :timeind :timefmt :textind
>  :deps :labels :xlabels :ylabels :xmin :xmax :ymin :ymax :ticks
>
> if they are customized by user in `org-plot/preset-plot-types'.

I don't follow your conclusion since this change will only affect the
value of :check-ind-type leaving the rest of the settings unaffected.

> I believe that the right way to address the problem will be
> `org-combine-plists' on the (1) org-plot/preset-plot-types; (2)
> org-plot/gnuplot-default-options; (3) #+PLOT lines in the buffer.

Looking at the definition of org-plot/add-options-to-plist and the Info
manual, I am not sure if :check-ind-type is supposed to be customised by
the PLOT line.  It seems to be more of an internal setting.  If you
agree to this, I can change the check to

    (when (or (plist-get type :check-ind-type) (plist-get params :check-ind-type))

to heed the value of :check-ind-type in org-plot/gnuplot-default-options
(since PARAMS has the default values included earlier in the defun).

>> [...]
>> The other code smell I see is that the function checks for the PLOT line
>> twice.  Once near the beginning of the function, and once just after the
>> cleaning up of hline.  Is this simply an oversight?
>
> It is kinda intentional, but broken.
>
> Historically, users can put #+PLOT lines _after_ the table.
> However, after refactoring org-table.el, this is no longer working.
> See https://list.orgmode.org/orgmode/87o7a0p9ba.fsf@localhost/

OK, I will leave the check in then.

>> Coming to the grid example, the doc-string of org-plot/preset-plot-types
>> options says:
>>
>>     - :data-dump - Function to dump the table to a datafile for ease of
>>       use.
>>
>>       Accepts lambda function.  Default lambda body:
>>       (org-plot/gnuplot-to-data table data-file params)
>>
>> but in fact, org-plot/gnuplot passes one more argument to the :data-dump
>> function:
>>
>>       ;; Dump table to datafile
>>       (let ((dump-func (plist-get type :data-dump)))
>>         (if dump-func
>> 	    (funcall dump-func table data-file num-cols params)
>> 	  (org-plot/gnuplot-to-data table data-file params)))
>>
>> but here's the catch: the :data-dump function in the grid option expects
>> the order
>>
>>     (lambda (table data-file params _num-cols)
>>
>> which breaks things down the line.  What should be the actual order
>> here?  I looked at the history of those lines briefly using C-x v h but
>> I don't have the time to look into it properly to decide on the actual
>> argument order.
>
> The best order is de-facto calling convention in the code:
>
> (funcall dump-func table data-file num-cols params)
>
> The docstring and the default value should be fixed accordingly.

OK, will do this in a future patch.