Make Text use Pango by default

[behackl]

List of Changes

This changes our default text renderer from the Cairo toy API to Pango. A logger.info message is added that points to the former implementation in the form of CairoText.

Testing Status

Tests pass (at least locally).

Acknowledgement

• I have read the Contributing Guidelines

[behackl]

Note: I put up this PR as a suggestion. I think it makes sense to switch to rendering with Pango by default. If there are people against doing so already, feel free to speak up; I don't mind closing this PR again (or leaving it open until we feel that it is a better time to switch).

[naveen521kk]

Thank you for picking this up✌.

[behackl]

If someone wonders about the RTD build time: the first time I pushed it we have 506 seconds (which is about right), and the second time (after removing an additional newline) 765 seconds.

I think the latter build is just an outlier, no need to worry about it.

[naveen521kk]

LGTM!

[eulertour]

I'm a little uncomfortable doing this because I still don't understand the new problem with Cairo installation on Windows.

Am I understanding correctly that pycairo and pangocairocffi both contain Cairo wheels, and that the two are incompatible with each other? And if that is the case, should we be working towards replacing pycairo entirely?

[behackl]

I'm a little uncomfortable doing this because I still don't understand the new problem with Cairo installation on Windows.

Are you referring to the problems where the import of pangocairocffi._generated.ffi fails? I might have missed other reports of installation failures.

Anyways, if so I think this is a problem with how pangocffi and pangocairocffi are installed and not necessarily an interaction problem between different cairo wheels -- but @naveen521kk has investigated this closer, I think?

Am I understanding correctly that pycairo and pangocairocffi both contain Cairo wheels, and that the two are incompatible with each other? And if that is the case, should we be working towards replacing pycairo entirely?

The description of cairocffi mentions that it is a drop-in replacement for pycairo; I don't know whether it makes sense to actually replace it.

[naveen521kk]

pycairo is different from all other things and not related to this.

Now, here I was forced to use cffi binding for pango as I couldn't find a CPython binding for it. So, I had to make it compatible as it is not possible to have connections between pycairo and cffi binding of Pango aka, pangocairocffi and pangocffi. So I had to cairocffi here. All these cffi binding requires interaction with binaries.

But the problem is users doesn't download the correct binaries, choco helps here. I had been try to help those projects to have self included wheels just like pycairo.

[eulertour]

In that case can we open an issue to get chocolatey working again?

[naveen521kk]

In that case can we open an issue to get chocolatey working again?

I did my best there. Just waiting for their mods to come up soon. Maybe a week?

[naveen521kk]

@behackl can you fix this function?

manim/manim/mobject/svg/text_mobject.py

Lines 771 to 783 in d6f7653

	def text2hash(self):
	"""Internally used function.
	Generates ``sha256`` hash for file name.
	"""
	settings = (
	"PANGO" + self.font + self.slant + self.weight
	) # to differentiate Text and PangoText
	settings += str(self.t2f) + str(self.t2s) + str(self.t2w)
	settings += str(self.line_spacing) + str(self.size)
	id_str = self.text + settings
	hasher = hashlib.sha256()
	hasher.update(id_str.encode())
	return hasher.hexdigest()[:16]

Remove that "PANGO" from the hash.

[behackl]

Remove that "PANGO" from the hash.

Wouldn't this cause CairoText and PangoText to produce the same hashes? I think it should be kept around as long as we keep CairoText around as a fallback.

[naveen521kk]

ok then it can be merged.