Re: [Gnuplot-info] pdfcairo: superscript in legend shifts text baseline

[Tatsuro M. <tma...@ya...>]

> From: Elias Assmann

> To: Tatsuro MATSUOKA; gnuplot-info
> Cc: 
> Date: 2015/5/24, Sun 05:30
> Subject: Re: [Gnuplot-info] pdfcairo: superscript in legend shifts text baseline
> 
> On 05/22/2015 01:17 PM, Tatsuro MATSUOKA wrote:
>>  I have tested on pdfcairo, pngcairo and wxt terminal on gnuplot 5.1
>>  (cvs, 2015-05-19) (windows version) and upload a zip archived file.
>> 
>>  http://www.geocities.co.jp/tmgpltwin/Files/Files.html#0064
>> 
>>  0064 test.zip, 105,699 bytes, 2015-05-22, zip archived test files on
>>  cairobased terminals on gnuplot 5.1
> 
> Thank you.  Meanwhile, I have also done some more analysis,
> using three variables: terminal (png/pdf), font (default font /
> Carlito), and gnuplot version (4.6 / 5.0).  Carlito
> <http://www.tug.dk/FontCatalogue/carlito/> is the font I use in the plot
> that originally led me to post.  With wxt, I get the same results as
> with pdfcairo.
> 
> First, in the png terminal I do not see the effect.  All the legend
> entries are identically vertically aligned w.r.t. their lines.  (Compare
> nosuper.png and default-5.0.png.)
> 
> With the pdf terminal, the superscripted legend entry is placed further
> down.  The space between legend lines is not changed.  (Compare
> nosuper.pdf and default-5.0.pdf.)  The shift is larger in 4.6 than in 5.0.
> 
> The effect also seems larger with Carlito than with the default font. What is 
> more, with the default font, the legend text (without superscripts) is 
> approximately centered vertically w.r.t. the line.  But for some reason, when 
> using Carlito, it is considerably off-center.
> 
> In your plots, I see the same behavior in the pdf and wxt versions, but
> curiously enough, the shift also happens in your png version.
> 
> You can find my plots at 
> <ftp://ftp.tugraz.at/outgoing/assmann/baseline.zip>:
> 
> * {carlito,default}*: plots as mentioned above
> * ruler-*: same, with rulers added for clarity
> * nosuper.*: without any superscripts
> * optcond_origu.pdf: this is my “real” plot; I think this is actually a better 
> illustration of the problem than the simple test


Elias


I did not understand what you mean correctly.
OK. The same behaviors should be expected on the different terminals concerning baseline.

Please file to bug ticket with the files you prepared.
The bug ticket allows attachment file. 

> In your plots, I see the same behavior in the pdf and wxt versions, but
> curiously enough, the shift also happens in your png version.


It is not curious. As I wrote, I have made png image by pngcairo terminal but not png terminal.

set term png => png terminal based on libgd
set term pngcairo => pngcairo terminal based on libcairo and libpango.

Note that "set term pdf" behaves in special way. 
1. If the original pdf terminal based on pdflib exists, "set term pdf" points to the original one.
2. If the original pdf terminal based on pdflib does not exist but the pdfcairo terminal exist,
"set term pdf" means the same as "set term pdfcairo".

The gnuplot windows binaries distributed on the SourceForge and on my personal site (5.1 cvs)[1], the original pdf terminal does not exist.
Then "set term pdf" is the same as "set term pdfcairo".

This is because the free version of pdflib is not easy to obtained in current situation.

 [1] http://www.tatsuromatsuoka.com/gnuplot/Eng/winbin/

Tatsuro