Indic Script labels not rendered correctly

[aquiire]

Bug report

Bug summary

Title sums it up all I guess.
Related:

• Issue #3517 on Github
• This and this on stack exchange

Here I used sinhala as Indic Script instance.
I think the font family has to do something with this issue since output to different font families are different. I'm not an expert in these stuff. Help appreciated to to get this work even with a different font family.
An example is given below for which I've tried with 3 different font families.
Code for reproduction

from matplotlib import rc
families = ['WARNA','LKLUG','Arial']
for familyname in families:
    rc('font', family=familyname)
    from matplotlib import pyplot as plt
    import numpy as np
    r = np.linspace(0.1,5,100)
    A = 3*2/(r*2)+np.pi*r**2
    fig = plt.figure()
    axes = fig.add_axes([.1,.1,.8,.8]) 
    axes.plot(r,A)
    axes.set_ylabel(u'වර්ගඵලය')
    axes.set_xlabel(u'අරය')
    axes.set_title(familyname)

Actual outcome

Expected outcome

Result 1 looks fine but it has 2 problems:

• Accents not been placed in the correct location. ("්" in "වර්ගඵලය" is placed after "ර" not before)
• Letter "A" is not renderd in the title "WARNA"

Issues in result 2:

• Same previous accent issue
• Axis values or title not shown at all

Issues in result 3:

• No Sinhala script shown.

Versions

Software	Version
Python	2.7.13 64bit [GCC 4.4.7 20120313 (Red Hat 4.4.7-1)]
IPython	5.1.0
OS	Linux 4.4.0 78 generic x86_64 with debian jessie sid
matplotlib	1.5.1

(Matplotlib from anaconda bundle)

Wed Jun 14 22:54:11 2017 +0530

[anntzer]

The misplaced accent is due to lack of support for combining characters at the freetype level (e.g. https://stackoverflow.com/questions/9985340/how-to-render-combined-unicode-characters-in-freetype). Using a real text layout engine (MEP14) is probably not going to happen any time soon :-/

Not sure what's happening with the missing A (well I guess the glyph is missing? unlikely though) and the missing title of case 2.

Case 3 is just that Arial does not provide Indic characters (in all likelihood) so you're out of luck.

[anntzer]

/self-promotion/ mplcairo now supports complex text layout using libraqm, this was suprisingly easy to add...

[baabsa]

can you please illustrate steps for adding and use (libraqm),

I am new in python and I didn't find an easy explanation how it is used.

Thanks in advance.

[anntzer]

It is a fairly technical project and will probably not be implemented any time soon.
But most solutions that fixes indic script will likewise fix arabic script (*well technically you can have a shaper without RTL support but let's not get there).

[baabsa]

this issue doesn't appear in printing an Arabic word from DataFrame, it is only appear when I have Arabic words in plot as x axes, y axes, or title!

[anntzer]

Actually this issue is basically already tracked at #5941, so I'll close this one as a duplicate.