After using subset_fonts, some Greece letters are not displayed

[izerui]

def test_subfont0():
doc = pymupdf.open()
page = doc.new_page(width=1024, height=768)
page.insert_htmlbox(
rect=pymupdf.Rect(280, 50, 480, 100),
text=f'Α、Β、Γ、Δ、Ε、Ζ、Η、Θ、Ι、Κ、Λ、Μ、Ν、Ξ、Ο、Π、Ρ、Σ、Τ、Υ、Φ、Χ、Ψ、Ω。α、β、γ、δ、ε、ζ、η、θ、ι、κ、λ、μ、ν、ξ、ο、π、ρ、ς、τ、υ、φ、χ、ψ、ω。'
)
doc.subset_fonts()
doc.ez_save('subseted.pdf')

[JorjMcKie]

If you do things like the above, you leave your fate in the hands of MuPDF's automatic font selection algorithm. If I stood in your shoes I would exert some influence and specify at least a font family to choose from, e.g. "sans-serif".
Other than that, why did you take "、" instead of a normal comma? That character, has the Unicode 0x3001 (12289, 'ideographiccomma') and is far outside the Extended Latin character set.
The same is true for "。" (0x3002, 12290, 'ideographicperiod'). Why not a normal colon?
Both characters only appear to be followed by a space, but there is no space, so crazy line breaks (in the middle of a word) can easily occur.

So with this text, you will end up with 3 fonts automatically identified by MuPDF. One of them, because of the 2 ideographic punctuations, is very large:

(6, 'otf', 'Type0', 'Noto Serif Regular', 'F0', 'Identity-H', 24)
(12, 'cid', 'Type0', 'Charis SIL Regular', 'F1', 'Identity-H', 24)
(18, 'otf', 'Type0', 'Source Han Serif TC Regular', 'F2', 'Identity-H', 24)

If you do page.insert_htmlbox(..., css="* {font-family: sans-serif;}"), it will be just the following two:

(6, 'cid', 'Type0', 'Nimbus Sans Regular', 'F0', 'Identity-H', 18)
(12, 'otf', 'Type0', 'Source Han Serif TC Regular', 'F1', 'Identity-H', 18)

If you replace the punctuation by normal comma and colon, only one front will be used (Nimbus Sans Regular).
Other than the above, there indeed is a problem with MuPDF's font subset algorithm, which I will report in their issue management.
You can use doc.subset_fonts(fallback=True) which will use the mechanism of the fontTools package.

[izerui]

doc.subset_fonts(fallback=True) it`s ok.

Thank you very much!!!

The only drawback is that the execution speed of subset_fonts is 4 to 5 times slower.