DEV Community: Mansour Moufid

Unicode bug on macOS

Mansour Moufid — Sun, 30 Apr 2023 18:13:37 +0000

A diacritical mark is a mark on letter that conveys meaning like a change in pronunciation, like an accent. Diacritical marks in Unicode can have one or more encodings.

For example the letter c with a cedilla (ç) can be the single character:

Character	UTF-8 encoding	Name
ç	0xc3 0xa7	LATIN SMALL LETTER C WITH CEDILLA

or as two:

Character	UTF-8 encoding	Name
c	0x63	LATIN SMALL LETTER C
	0xcc 0xa7	COMBINING CEDILLA

The first is known as "normal form C" (NFC). The second is called "normal form D" (NFD) -- it is the canonical decomposition of normal form C.

Something interesting happens when you save a file with a name in NFC: macOS will not open it.

For example:

filename = 'français.txt'
with open(filename, 'wt') as f:
    print('Bonjour!', file=f)

Try to double-click this file in Finder. TextEdit will show in the Dock, but nothing else happens.

Now close TextEdit, delete the file, and change the file name to its canonical decomposition (NFD):

import os
import unicodedata

filename = 'français.txt'
filename = unicodedata.normalize('NFD', filename)

try:
    os.remove(filename)
except FileNotFoundError:
    pass
with open(filename, 'wt') as f:
    print('Bonjour!', file=f)

TextEdit will open this file just fine.

If you're wondering why a bunch of your files stopped opening after the latest upgrade to macOS (13.3.1), this is why...

Make a macOS app icon from an image

Mansour Moufid — Wed, 15 Mar 2023 13:06:31 +0000

macOS has everything you need to make an icon file for your application from the terminal.

macOS application icons are really a bundle ("iconset") of images at various sizes (16×16, 32×32, 64×64, ..., 1024×1024).

First create a myicon.iconset directory containing your icon image at all the sizes needed using the sips command line tool:

sips -Z 16 myicon.png --out myicon/icon_16x16.png
sips -Z 32 myicon.png --out myicon/icon_16x16@2x.png
sips -Z 32 myicon.png --out myicon/icon_32x32.png
sips -Z 64 myicon.png --out myicon/icon_32x32@2x.png
...

Then bundle them together with the iconutil command:

iconutil -c icns myicon.iconset

Here is a Makefile to do this automatically. Just save the Makefile to a directory containing your icon image file; change the variables NAME and ICON at the top; type make in the Terminal, and you're done: https://gist.github.com/eliteraspberries/a536f28b0498ec844fb210bb597738e2

Quick bytes to integer (or other type) conversion in Python

Mansour Moufid — Tue, 20 Dec 2022 00:33:35 +0000

Sometimes you have a bytes object but what you really want is an integer.

For example, I want the unsigned 32-bit integer represented by b'1234'. I can do the conversion with struct.unpack:

>>> struct.unpack('I', b'1234')
(875770417,)

or using the from_buffer/from_buffer_copy functions of the ctypes module:

>>> ctypes.c_uint32.from_buffer_copy(b'1234')
c_uint(875770417)

This is usually used for compound types (also called composite data types, i.e. types that are defined in terms of primitive data types, like a C structure), but still useful for simple things like integers.

Programmatically create submodules in Python

Mansour Moufid — Tue, 11 Oct 2022 21:49:14 +0000

If you've ever wanted to distribute a Python package as a single file, but keep the benefits of dotted namespaces, here is a tip.

We can generate submodules programmatically and add them to the current module's namespace using importlib.

For example, here is a single file module (named x if you save it to a file named x.py) with a submodule named y:

import importlib.machinery
import importlib.util
import sys

spec = importlib.machinery.ModuleSpec('y', None)
y = importlib.util.module_from_spec(spec)
sys.modules[__name__].y = y

y.abc = 123

In order to do import x.y or from x import y as if x were a package and not simply a module, just add:

sys.modules[__name__ + '.y'] = y

Encode and decode FourCC codes, in Python

Mansour Moufid — Mon, 28 Mar 2022 23:37:33 +0000

Convert the string "H264" to an integer code:

sum([ord(c) << (i * 8) for i, c in enumerate("H264")])

Convert the integer code (0x34363248) back to a string:

''.join([chr((0x34363248 >> (i * 8)) & 0xff) for i in range(4)])

See this gist: https://gist.github.com/eliteraspberries/b017f998c97c395881f0d2f05a08ec60

Python unique list, in order

Mansour Moufid — Tue, 07 Dec 2021 02:54:55 +0000

Do you need to remove duplicates from a list, and keep it in order?

xs = [1, 3, 2, 2, 5, 4, 1, 5]

The set type will return one of each element:

>>> set(xs)
{1, 2, 3, 4, 5}

But not in order, because a set is an unordered collection.

>>> for x in set(xs):
...     print(x)
... 
1
2
3
4
5

Instead, first construct a set from the list, then a list from this set, in the right order with the sorted function, like so:

>>> xs = [1, 3, 2, 2, 5, 4, 1, 5]
>>> list(sorted(set(xs), key=lambda x: xs.index(x)))
[1, 3, 2, 5, 4]

Quiz: When does the above construct not work?