serhii.net

Saw this in the python pandoc cookbook¹

holder[index:index+1] = split_home(elt)

Wow.

Never thought I could assign multiple elements to a slice!

pdbpp is a drop-in replacement for pdb, and I like it more than ipdb for some reason.

Installing it makes it the default one imported when importing pdb (incl. by pytest, python’s breakpoint() etc!)

Really nice tutorial: pdb++, a drop-in replacement for pdb (the Python debugger) | PythonRepo

Vanilla-pdb cheatcheet: Python Debugger Cheat Sheet - Kapeli

Features not present in pdb that I love:

• ll outputs the text of the current function
• sticky updates the function listing with each new line, giving a nice interactive visual feeling to the debugging process

pytest -s works to make it play nice with the stdouts generated by pdbpp.

In the context of reading a settings.ini from python’s decouple¹ config lib, this works as empty string

YAML_CONVERTER_PREFIX=

has to be cast to string though:

D_YAML_CONVERTER_PREFIX = config("YAML_CONVERTER_PREFIX", cast=str)

These don’t, these are strings containing two characters, "" and '' respectively.

YAML_CONVERTER_PREFIX=""
YAML_CONVERTER_PREFIX=''

Wooho!

files = list(input_dir.glob("*.md"))[: cs.limit]
if output_path.is_file() and ((l := len(files)) != 1):
    raise ValueError(f"Can't write {l} files to single file {output_dir}")

Had to use additional parentheses around the actual assignment. Without that, black fails in an interesting way:

error: cannot format smw_to_hugo/yaml_converter.py: cannot use --safe with this file; failed to parse source file.

Just discovered this! In vim, if I skip the pattern, it’ll take the one last searched for:

/mypattern
:s//newpattern/g

Had weird issues with kitty terminal output being wrong, lines in vim/nvim being in wrong places, usually because it thought the terminal was a different size than it really was (blamed it on nvim initally, but the problem happened in other complex CLI programs too, notably tig).

$TERMINFO wasn’t set, and the terminfo file was nowhere to be found. The package kitty-terminfo was installed though.

In any case, downloaded the terminfo file from the repo and set the env variable manually in zshrc, now it works:

export TERMINFO="$HOME/.config/kitty/xterm-kitty"

After for the nth time writing awkward code like

if limit is None: 
    limit = len(mylist)

decided to see if there’s a better way. Looked into the walrus operator etc,but decided to test what I get with None.

Well, mylist[:None] works! No errors, I’d guess I get a copy of it same as mylist[:].

Will save me hundreds of lines in the future!

Docu about slice¹ is terse, says it uses range(start,end,step) under the hood with start and step defaulting to None. But range doesn’t accept None for all arguments! TODO for later I guess.

Found a post¹ about it.

But I like much more Click’s way to do this (Options — Click Documentation (8.0.x)):

@click.option(
    "--username", prompt=True,
    default=lambda: os.environ.get("USER", "")
)

Of course, os.get.environ can be replaced by python-decouple’s config().

Lastly, ini files support interpolation² (%(whatever)s)! Final solution:

[settings]
EXPORT=../../exports
CATS_INPUT=%(EXPORT)s/cats.json

@click.option(
    "--input-file",
    "-i",
    type=click.Path(exists=True, path_type=Path),
    default=lambda: config("CATS_INPUT"),
)

Also TIL if I use quotes in the ini file, they’ll become part of the final filename.

Stumbled upon python-decouple · PyPI, which seems like a “better” dotenv (supports casting, defaults etc)

For example, this is a settings.ini in poetry project root:

[settings]
ECHO=True

I can overwrite these parameters like ECHO=False poetry run python myscript.py

Neat!

Python Best Practices for a New Project in 2021 - Alex Mitelman

Describes a setup that uses poetry, black, flake8, pytest, mypy and new to me isort to sort imports.

The Fast track section has a TL;DR of how to create that setup.

I also really like this intro to poetry: Package Python Projects the Proper Way with Poetry