Several x.text angle and size problems problems

[jypeter]

I have found several serious problems when experimenting with x.text, that can be reproduced with the following simple script

import vcs
x = vcs.init()

x.text('String plotted with "y.text(''This string'',.05,.5,color=241,angle=0)"',.05,.5,color=241,angle=0)

x.text('Testing the text angle with +45.....',.5,.5,color=242,angle=45)

x.text('Testing the text angle with -45.....',.5,.5,color=243,angle=-45)

x.png('text_angle_sign_pb')
x.pdf('text_angle_sign_pb')

• The sign of the text angle is wrong, in both 2.1.0 and 1.5.1! In the png below, ngle +45 should be pointing to the upper right corner, not the lower right... It is also wrong in the Gallery example http://uvcdat.llnl.gov/display.php?file=test_basic_text . I had some doubts about my sanity and I have asked several people and checked on wikipedia, in case the US and European conventions were different: positive rotations are anticlockwise (http://en.wikipedia.org/wiki/Angle#Positive_and_negative_angles)
• The font size is different in the png output, between 2.1.0 and 1.5.1. 2.1.0 size is smaller, as
  can be seen in the png files below (2.1.0 is at the top)
  • 2.1.0
    
  • 1.5.1
    
• The angles of the 1.5.1 pdf are consistent with the 1.5.1 png file, but the angles of the 2.1.0 pdf are all ZERO. Note: the UV-CDAT logo is missing in the 2.1.0 pdf...
• 2.1.0 pdf file: https://files.lsce.ipsl.fr/public.php?service=files&t=0e5ebcd2d07a494ddbecfb4348859796
• 1.5.1 pdf file: https://files.lsce.ipsl.fr/public.php?service=files&t=e484f72a3a43194f07d79f043ee5e05d
• The text size is not consistent between png and pdf for 2.1.0, as can be seen in the screenshot below. It was already not consistent in 1.5.1, but the difference was smaller. 2.1.0 comparison is at the top, and for each screenshot, pdf is at the top, and png at the bottom
  • 2.1.0
  • 1.5.1

Note: @chaosphere2112 can add this example or a variation of it to the Gallery, once the angles are OK (http://uvcdat.llnl.gov/gallery.php?primitive=text)

[alliepiper]

Several of the PDF issues are due to limitation is the PDF specification (believe it or not, PDF doesn't support text alignment flags, etc). This has been addressed in master by exporting all text as a set of bezier curves rather than text tags, as this will ensure equivalent output on-screen, in png, and in the vector backends.

Rotation direction is a matter of convention and differs across toolkits and domains. VTK uses the "positive = CCW" (as the trigonometry is more straightforward), but vcs appears to do the opposite when constructing the VTK text properties: p.SetOrientation(-to.angle). The VCS manual doesn't seem to specify their convention -- it specifies that "angel" (a typo?) is just an int between 0 and 360, though it seems to also accept negative angles as well from your example code.

The text size is a mystery to me. The manual specifies that text "height" is an integer between 1 and 100. I have no idea what units. @doutriaux1 wrote the bit that converts the VCS height to a fontsize in vcs2vtk.prepTextProperty -- perhaps a scale factor or similar needs tweaking?

In short -- The PDF issues are (hopefully) fixed once and for all in master. Vector exported text should be identical to what is shown on screen, as it is now explicitly drawn in the exported image (rather than leaving the glyph loading/placement up to the viewer). The absolute text height units and text rotation conventions are documentation concerns for vcs -- @doutriaux1, can you find out what these conventions are supposed to be?

@jypeter Thanks for the very thorough testing and report!

[jypeter]

• wow, no text alignment in pdf? I would have supposed that it was a basic requirement... Text as curves seems like an awful overkill. And then forget importing vcs pdf in Illustrator to make some easy text changes
• The positive = CCW convention is what any sane person who has followed trigo classes will expect, so vcs should not break that. Maybe it's an old vcs bug (since it was already in 1.5.1). Can this be fixed without breaking too many things? I'm happy with not being backward compatible with old scripts if we fix that
• There is a good chance that the text size is proportional to the canvas size. We have been dealing with canvas coordinates between 0 and 1 since last century and we are probably stuck with this. But I think it's a good thing if the text changes size when the vcs canvas is resized

[chaosphere2112]

Yup, text size is proportional to the height of the window; the ratio currently is text.height * window height / 800, rounded down to an integer.

[durack1]

@jypeter @dlonie @chaosphere2112 @doutriaux1 again another related to #503

[alliepiper]

Yes, as complex as the PDF specification is, text support lacks some basic options ;) It's indeed a shame to lose the ability to edit the text directly, though it's simple enough to just replace the path with a new text object in an editor if needed. The path approach also has the advantage of ensuring consistent output across platforms/viewers, which often load different fonts for various face types and can cause some real headaches with alignment.

[jypeter]

Well, as long as it works and indeed everything is clean and look the same...

Devil is in the details and my end users can sometimes be very picky about some details (change the color, move this over there, align the longitudes differently, etc...). So if I get something right on screen, it's better if I get the same in the output

[doutriaux1]

@jypeter the sign of the angle is following XGKS specification I agree it's not ideal but it is to preserve backward compatibilty. We can discuss "breaking" this but I really don't like the idea.

[durack1]

@doutriaux1 as the last update to XGKS was in 2000, not sure this is a valid reason this decade, let alone century..

Right now is also an opportune time to break some backward compatibility, as with a significant rewrite comes change no matter how you like it, and it hurts less if it's all in one hit..

[alliepiper]

Closing this, as the issues have been resolved as follows:

• Angle sign conventions AIUI, these are consistent with previous versions of VCS.
• Angle/size issues in PDF exports Fixed by exporting all text as bezier paths to work with the PDF feature palette and prevent platform-dependent text issues.

If there are any unresolved problems from this thread, let's create new issues for them so we can more easily keep track of the individual tasks.