Python Types for Union of String Literal and Other Type

I wanted to add a single special value to another type for a union type, something like this:

from typing import Literal
                                                                            
UNMEASURED = "unmeasured"                                                    
Unmeasured = Literal[UNMEASURED]
Resource = int | Unmeasured

notes: dict[Resource, str] = {1: "foo", 2: "bar", 3: "baz", UNMEASURED: "quux"}
resources = [1, 2, 3] + [UNMEASURED]
for resource in resources:
    print(notes[resource])

This works, but mypy complains: ‘error: Parameter 1 of Literal[…] is invalid [valid-type]’. So to make this pass mypy typechecking you have to repeat “unmeasured” as a literal. Even if you do that:

from typing import Literal

UNMEASURED = "unmeasured"
Unmeasured = Literal["unmeasured"]
Resource = int | Unmeasured

notes: dict[Resource, str] = {1: "foo", 2: "bar", 3: "baz", UNMEASURED: "quux"}
resources = [1, 2, 3] + [UNMEASURED]
for resource in resources:
    print(notes[resource])

mypy will complain: ‘Dict entry 3 has incompatible type “str”: “str”; expected “int | Literal[‘unmeasured’]”: “str” [dict-item]’ and also ‘Invalid index type “str | int” for “dict[int | Literal[‘unmeasured’], str]”; expected type “int | Literal[‘unmeasured’]” [index]’

Eventually I came up with:

from enum import Enum
from typing import Literal

class Aspect(Enum):
    UNMEASURED = "unmeasured"

Resource = int | Aspect

notes: dict[Resource, str] = {1: "foo", 2: "bar", 3: "baz", Aspect.UNMEASURED: "quux"}
resources = [1, 2, 3] + [Aspect.UNMEASURED]
for resource in resources:
    print(notes[resource])

which both produces the same result, and passes the typechecker. I don’t love the weird Enum hanging out on its own but it does at least put the magic string in one place only and pass. I wonder what would be better.

Remember the Milk Alfred Workflow

This uses rtm-cli which you can install with npm.

query=$1
logfile="/Users/bakert/u/scratch/rtm.log"
rtm=/opt/homebrew/bin/rtm
max_lines=1000

# Log the query
echo "[$(date)] $query" >> "$logfile"

# Ensure the logfile does not grow beyond $max_lines
tail -n $max_lines "$logfile" > "$logfile.tmp" && mv "$logfile.tmp" "$logfile"

# Send the task to Remember the Milk
"$rtm" add $query

if [ $? -ne 0 ]; then
  echo "SOMETHING WENT WRONG! Please check that rtm-cli is installed: npm install -g rtm-cli"
else
  echo "[Added] $query"
fi

Gmail Conversations Archiving Has Gotten Worse

You can search for emails in Gmail, “Select All” and Gmail will prompt you, “do you want to select all conversations that match this search?” rather than just the 100 you can see. Then if you do something like use the Archive keyboard shortcut it says “do you want to apply this to all conversations selected?” Previously this was instant. Now it says, “We archived some conversations. We’ll do the same for any remaining conversations in a few minutes. This might take longer, depending on how many conversations are selected.” And if you go back to your Inbox they’re still there. A real downgrade!

assertRaises in table-driven tests

This is what I want to do but it fails passing None to assertRaises:

tests = [
   (0, None),
   (1, None),
   (-1, TooFewException),
   (99, None),
   (100, TooManyException),
]
for n, exc in tests:
    with self.assertRaises(exc):
        results = my_code(n)
        assert len(results) == n

Here’s a version of assertRaises that will let you do that:

    def assert_raises(self, exception: Type[Exception]):
        if exception:
            return self.assertRaises(exception)
        return contextlib.nullcontext()

Line Count

Number of lines in a (python + js) codebase:

#!/bin/zsh

# This script counts the number of non-blank lines of code in a directory and its subdirectories
# for Python and JavaScript code only

# Set the directory to search for code files
dir='.'

# Count the number of non-blank lines of code for Python and JavaScript files
num_lines=$(find "$dir" -type f \( -name '*.py' -or -name '*.js' \) -not -path '*venv*' -not -path '*node_modules*' -exec grep -he '^[^[:space:]]' {} + | awk 'NF{count++} END{print count}')

echo "Number of non-blank lines of code: $num_lines"

Number of lines per-file in a (python + js) codebase:

#!/bin/zsh

# This script counts the number of non-blank lines of code in a directory and its subdirectories
# for Python and JavaScript code only

# Set the directory to search for code files
dir='.'

# Count the number of non-blank lines of code for Python and JavaScript files
num_lines=$(find "$dir" -type f \( -name '*.py' -or -name '*.js' \) -not -path '*venv*' -print0 | xargs -0 grep -cve '^[[:space:]]*
)

echo "Number of non-blank lines of code: $num_lines"

Queries on local so much faster than queries on server

I was encountering a situation where big aggregating queries that took seconds on local were taking minutes or even hours on prod. After trying about a million things this Stack Exchange post finally cleared things up.

Setting innodb_buffer_pool_size = 5G in MariaDB’s configuration and restarting instantly changed the queries to taking seconds. The default is 128M which is … not useful 🙂

Change Terminal Background Color When ssh’ed – Mac

Save this as “ssh” somewhere on your $PATH ahead of /usr/bin/ssh and make it executable

#!/bin/sh

HOSTNAME=$(echo $@ | sed 's/.*@//')

set_theme () {
    osascript -e "tell application \"Terminal\" to set current settings of first window to settings set \"$1\""
}

on_exit () {
    set_theme "Solarized Light"
}

trap on_exit EXIT

case $HOSTNAME in
  # These all do the same thing but we could have different themes for different locations
  hb|pd|decksite) set_theme "Solarized Dark" ;;
  *) set_theme "Solarized Dark" ;;
esac

/usr/bin/ssh "$@"