DEV Community: Benji 🍙

My sudden episode to keep my virtualenv in check

Benji 🍙 — Thu, 08 Jan 2026 13:37:04 +0000

I often run python scripts or commands outside of a virtualenv and this creates a problem with dependencies being installed globally, often left unchecked with wrong python versions.

The solution is simple; If i'm not running a virtualenv anywhere then ideally I should get a warning message, loudly. What i've ended up doing is creating a wrapper in my zsh config file that looks something like below:

# Python wrapper to warn about uv venv
python() {
environment
    if [ -z "$VIRTUAL_ENV" ] || [ ! -f "$VIRTUAL_ENV/pyvenv.cfg" ] || ! grep -q "uv" "$VIRTUAL_ENV/pyvenv.cfg" 2>/dev/null; then
        echo "\033[1;31m" >&2  # Bold Red
        echo "" >&2
        echo " ██╗    ██╗ █████╗ ██████╗ ███╗   ██╗██╗███╗   ██╗ ██████╗ ██╗" >&2
        echo " ██║    ██║██╔══██╗██╔══██╗████╗  ██║██║████╗  ██║██╔════╝ ██║" >&2
        echo " ██║ █╗ ██║███████║██████╔╝██╔██╗ ██║██║██╔██╗ ██║██║  ███╗██║" >&2
        echo " ██║███╗██║██╔══██║██╔══██╗██║╚██╗██║██║██║╚██╗██║██║   ██║╚═╝" >&2
        echo " ╚███╔███╔╝██║  ██║██║  ██║██║ ╚████║██║██║ ╚████║╚██████╔╝ ██╗" >&2
        echo "  ╚══╝╚══╝ ╚═╝  ╚═╝╚═╝  ╚═╝╚═╝  ╚═══╝╚═╝╚═╝  ╚═══╝ ╚═════╝  ╚═╝" >&2
        echo "" >&2
        echo "\033[1;33m UV virtual environment is NOT activated!\033[0m" >&2
        echo "\033[0;36m Run: uv venv && source .venv/bin/activate\033[0m" >&2
        echo "" >&2
        return 1
    fi

    # Execute the actual python command with all arguments
    command python "$@"
}

The above wrapper is specifically designed for uv-managed environments, checking not just if $VIRTUAL_ENV is set but also validating that it's actually a uv environment by grepping the pyvenv.cfg file.

Whenever i forget to activate my virtualenv, i'll get a massive ASCII art "WARNING!" banner in bold red, followed by instructions in yellow and cyan. It's obnoxious in the best possible way—you literally cannot miss it.

IMO the beauty in this is the Docker-esque approach so anytime I'm done with an environment, or I want to use a different Python version I can just deactivate and remove the current .venv module and start again.

Sometimes the best development tools are the ones that yell at you when you're about to do something really dumb or just out of impulse.

TIL about the command `nohup`

Benji 🍙 — Tue, 23 Dec 2025 15:48:03 +0000

TIL about the command nohup, which stands for "no hang up". It's a command that runs a process immune to hangup signals (SIGHUP).

When you close your terminal or log out then the shell normally sends a SIGHUP signal to all running processes which terminates them. nohup prevents this and allows processes to continue running even after you've close the terminal.

I particularly needed this because I was locally running an n8n workflow, and although there is a Docker container version, I didn't want the hassle in adding more containers that were already running and so if I could somehow find a way to run it as a daemon then that would be perfect. This was when I discovered nohup exists and is already installed by default on my macbook.

To use it (like a background Docker container), you use the command similar to below:

nohup command [arguments] > /dev/null 2>&1 &

This means, run command [arguments] and any stdout/stderr just gets shoved into /dev/null (as my friend puts it, a rubbish bin where anything sent in is discarded ... like rafas opinions). The 2>&1 just means to redirect stderr to same place as stdout, and finally the last & is to run in the background

So in the end I ended up with the command:

nohup n8n start > /dev/null 2>&1 &

Reference: https://www.digitalocean.com/community/tutorials/nohup-command-in-linux

TIL you can output `aws ecr describe-images` as a table

Benji 🍙 — Mon, 09 Jun 2025 13:59:51 +0000

Whenever I want to describe an image of a container image I pushed on AWS ECR i would see a JSON similar to below:

aws ecr describe-images --repository-name <my-repo> --region <my-region> --query 'sort_by(imageDetails,&imagePushedAt)[-1]'

{
    "registryId": "*****",
    "repositoryName": "*******************",
    "imageDigest": "sha256:***************************************",
    "imageTags": [
        "****"
    ],
    "imageSizeInBytes": **********,
    "imagePushedAt": "***********************",
    "imageManifestMediaType": "***********************************************",
    "artifactMediaType": "***********************************************"
}

but with the suffix --output table you can get a (IMO) much better layout

aws ecr describe-images --repository-name <my-repo> --region <my-region> --query 'sort_by(imageDetails,&imagePushedAt)[-1]' --output table
-------------------------------------------------------------------------------------------------------
|                                           DescribeImages                                            |
+-------------------------+---------------------------------------------------------------------------+
|  artifactMediaType      |  application/vnd.docker.container.image.v1+json                           |
|  imageDigest            |  sha256:xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx  |
|  imageManifestMediaType |  application/vnd.docker.distribution.manifest.v2+json                     |
|  imagePushedAt          |  2025-01-01T00:00:00.000000+00:00                                         |
|  imageSizeInBytes       |  123456789                                                                |
|  registryId             |  000000000000                                                             |
|  repositoryName         |  my-repo                                                          |
+-------------------------+---------------------------------------------------------------------------+
||                                             imageTags                                             ||
|+---------------------------------------------------------------------------------------------------+|
||  my-tag                                                                                   ||
|+---------------------------------------------------------------------------------------------------+|

TIL you can apply git fixup automatically with git absorb

Benji 🍙 — Thu, 23 Jan 2025 10:47:55 +0000

One thing to be clear is that git absorb is not an official command , at least not with version 2.39.2 which is what I'm using.

The idea is, if you've ever used git fixup you'll likely be aware you have to do something like this:

git add <my file>
git commit --edit --fixup <commit SHA>

# And then rebase to keep your commit history clean and atomic
git rebase --autosquash --interactive origin/master

What this guy does, is it will automatically find the correct commit SHA for you, and automatically if you append --and-rebase

The result is your commit history looks and reads like a book:

commit 2: Feature B
commit 1: Feature A

instead of:

commit 3: linting issue for Feature A
commit 2: Fix bug for failing test for Feature A
commit 2: Feature B
commit 1: Feature A

TIL how to profile `zsh` start time

Benji 🍙 — Thu, 16 Jan 2025 10:43:21 +0000

Disclaimer: I'm heavily referencing from this blog post

Tl:Dr: Package managers like nvm, pyenv etc are initialized on startup and if you don't need one of them then just remove them from zshrc.

Context

Everytime I load a new session on iterm I notice a significant wait time, so out of curiosity I googled around and discovered how to profile the ./zshrc file by using the below steps:

❯ time zsh -i -c exit
zsh -i -c exit  0.87s user 1.13s system 42% cpu 4.685 total

The above command creates a temporary zsh subshell then measures the time it takes for the exit command to run within it. Finally, it prints the execution time to the console, which you see above is 4.6s

You can profile zsh by adding at the top zmodload zsh/zprof and zprof at the bottom. zprof is a built in profiler for zsh, and when executed I see the output similar to below:

num  calls                time                       self            name
-----------------------------------------------------------------------------------
 1)    6        2385.86   397.64   12.79%   2385.57   397.59   12.79%  __conda_activate
 2)    4        1959.50   489.88   10.50%   1959.50   489.88   10.50%  git
 3)   32        1940.77    60.65   10.40%   1940.77    60.65   10.40%  mcfly_prompt_command
 4)    6        1512.94   252.16    8.11%   1512.94   252.16    8.11%  compdump

Your output might show something else, as my personal laptop shows nvm being the culprit, while my work laptop shows the above shows that conda is taking up most of the time. Conda is a package manager tool similar to pyenv, which I barely use and only had when experimenting with something months back.

Anyway, looking through my zshrc file I spot this to be the reason for the wait:

125 # >>> conda initialize >>>
126 # !! Contents within this block are managed by 'conda init' !!
127 __conda_setup="$('/Users/benji/miniconda3/bin/conda' 'shell.zsh' 'hook' 2> /dev/null)"
128 if [ $? -eq 0 ]; then
129     eval "$__conda_setup"
130 else
131     if [ -f "/Users/benji/miniconda3/etc/profile.d/conda.sh" ]; then
132         . "/Users/benji/miniconda3/etc/profile.d/conda.sh"
133     else
134         export PATH="/Users/benji/miniconda3/bin:$PATH"
135     fi
136 fi
137 unset __conda_setup
138 # <<< conda initialize <<<

As I'm fearful I'll need it again one day, I decide not to remove it and opted to comment it out instead. I then source ~/.zshrc and see that the start time was reduced from 4.6s to 0.9s:

❯ time zsh -i -c exit
zsh -i -c exit  0.22s user 0.25s system 48% cpu 0.972 total

TIL that the `requests` library supports automatic retries with exponential backoff

Benji 🍙 — Fri, 06 Dec 2024 10:01:14 +0000

You can use a customized Adapter and force several retries with an exponential backoff factor on all HTTP/HTTPS requests. See example below:

import requests
from requests import adapters
from urllib3.util import Retry

# Create a transport adapter with a custom retry strategy.
retries = Retry(
    total=3,
    backoff_factor=3,
    status_forcelist=[500, 502, 503, 504]
)
adapter = adapters.HTTPAdapter(max_retries=retries)

# Ensure adapter is used for both HTTP and HTTPS requests.
session = requests.Session()
session.mount('https://', adapter)
session.mount('http://', adapter)

# Testing the retry mechanism
response = session.get("http://httpbin.org/status/500")

This returns the error below:

RetryError: HTTPConnectionPool(host='httpbin.org', port=80): Max retries exceeded with url: /status/500 (Caused by ResponseError('too many 500 error responses'))

The unfortunate thing is that there doesn't seem to be a way to tell how many times the above mechanism has attempted to retry, only when all attempts have been exhausted

Reference

https://stackoverflow.com/a/47475019/4477547

TIL how to see the entire commit column on GitLab using JS

Benji 🍙 — Wed, 27 Nov 2024 15:36:22 +0000

This was quite a while back but I'm adding this here as a note-to-self thing.

Whenever you open CI/CD > Pipelines on GitLab you might often see something like this:

You can see the whole thing by opening up the dev console on your browser and changing some CSS to fix the styling. Another way you can also see the entire column is by creating a new bookmark (on any page, it doesn't matter) and then edit the URL to be this:

javascript:(function(){var e=document.querySelectorAll('a.commit-sha.mr-0');for(var t=0;t<e.length;t++){e[t].parentNode.removeChild(e[t]);}var n=document.querySelectorAll('div.container-limited');if(n.length>0){n[0].style.maxWidth='2400px';}var r=document.querySelectorAll('a.ref-name');for(var t=0;t<r.length;t++){r[t].style.overflow='unset';}})();

So then when you click it, the table will resize to look like below:

TIL you can create keyboard shortcuts to switch between a specific desktop

Benji 🍙 — Sun, 30 Jun 2024 13:51:32 +0000

Normally on a mac, you can use the trackpad to swipe left or right to switch between a desktop. But you can add a shortcut to switch between desktop x instead of swiping multiple times, or swiping up to select one.

How to do this

 > System preferences > Keyboard > Keyboard shortcuts > Mission Control > Tick the checkbox for mission control > Click the caret symbol on the left and check all the ones you want to add to a shortcut.

Reference: https://qiita.com/saboyutaka/items/d6cfd2a2b60f1a374d60#%E3%83%87%E3%82%B9%E3%82%AF%E3%83%88%E3%83%83%E3%83%97%E3%82%92%E5%A2%97%E3%82%84%E3%81%97%E3%81%A6%E3%82%B7%E3%83%A7%E3%83%BC%E3%83%88%E3%82%AB%E3%83%83%E3%83%88%E5%89%B2%E3%82%8A%E5%BD%93%E3%81%A6

TIL you can restore deleted files from a commit and then apply a git fixup to update the changes of that commit instead.

Benji 🍙 — Fri, 17 May 2024 09:53:00 +0000

git checkout [previous_commit_hash]^ -- [file_path]
a. The ^just specifies the parent commit of the previous commit hash where the file shouldn't have been deleted
(optional) Edit your file if you want to include any change in the commit instead
git commit --edit --fixup [previous_commit_hash]
Rebase with autosquash so you apply the fixup commit to the previous commit hash git rebase --autosquash --interactive origin/master
Force push (with lease) to your feature branch git push origin [your_branch] --force-with-lease

Any changes in the previous commit are just retained as normal

The other way you can do this (if its the latest commit) is to just git reset HEAD~1 to undo those changes, checkout the file that was deleted so its no longer included, then commit again with the same commit message. You'd still need to force push though since you're rewriting the commit history

TIL you can re-order your commit history through rebase -i

Benji 🍙 — Tue, 13 Jun 2023 14:42:53 +0000

Suppose you have a commit history like this:

commit 5: Refactor code for better performance
commit 4: Add new feature: user authentication
commit 3: Fix typo in README file
commit 2: Update documentation with usage instructions
commit 1: Initial commit with project setup

And then you have a new commit "Another README for user authentication" that should go after commit 4 and before commit 5:

commit 6: Another README for user authentication
commit 5: Refactor code for better performance
commit 4: Add new feature: user authentication
commit 3: Fix typo in README file
commit 2: Update documentation with usage instructions
commit 1: Initial commit with project setup

Then using git rebase -i HEAD~x (where x represents the number of commits you want to include in the rebase) you can re-order the commit history with relative ease, like this:

git rebase -i HEAD~6

This opens up your editor as usual

pick 6: Another README for user authentication
pick 5: Refactor code for better performance
pick 4: Add new feature: user authentication
pick 3: Fix typo in README file
pick 2: Update documentation with usage instructions
pick 1: Initial commit with project setup

And moving the commit to where you want

pick 5: Refactor code for better performance
pick 6: Another README for user authentication
pick 4: Add new feature: user authentication
pick 3: Fix typo in README file
pick 2: Update documentation with usage instructions
pick 1: Initial commit with project setup

Save then close the editor and it's done. Confirm in your git log before pushing remotely

commit 6: Refactor code for better performance
commit 5: Another README for user authentication
commit 4: Add new feature: user authentication
commit 3: Fix typo in README file
commit 2: Update documentation with usage instructions
commit 1: Initial commit with project setup

TIL you can convert media file extensions with ffmpeg

Benji 🍙 — Sun, 28 May 2023 12:30:26 +0000

You can convert a video clips file extension using ffmpeg like this:

$ ffmpeg -i chill.mp4 chill.wav
$ ll
-rw-r--r--  1 hello  world  84317363 21 Jun  2022 chill.mp4
-rw-r--r--  1 hello  world  49074254 28 May 13:13 chill.wav

You can also play directly from the terminal (which opens in a new window)

$ ffplay chill.mp4

And probably the most useful I found, which is to convert a video clip to a .gif for PRs/MRs

ffmpeg -i in.mov -s 600x400 -pix_fmt rgb24 -r 10 -f gif - | gifsicle --optimize=3 --delay=3 > out.gif

Notes on the arguments:

-r 10 tells ffmpeg to reduce the frame rate from 25 fps to 10

-s 600x400 tells ffmpeg the max-width and max-height

--delay=3 tells gifsicle to delay 30ms between each gif

--optimize=3 requests that gifsicle use the slowest/most file-size optimization

(full credits for this one go to this README.md)

TIL you can trick Mypys typechecker using typing.cast

Benji 🍙 — Wed, 24 May 2023 10:29:30 +0000

Problem

You might have something like this:

# models.py
class MyModelQuerySet(QuerySet[MyModel]):
    # ...

# views.py
class MyView():
    def get_queryset(self) -> QuerySet:
        # ...
        qs = self.model.stuff()

        # ...
        qs = need_other_stuff.get_other_stuff(queryset=qs)

And when you run Mypy it will spit out the below error:

Incompatible types in assignment (expression has type "_QuerySet[MyModel, MyModel]", variable has type "MyModelQuerySet[MyModel]")  [assignment]

This error points to the following lines:

# ...
qs = self.model.stuff()
# ...
qs = need_other_stuff.get_other_stuff(queryset=qs)

You could explicitly type annotate them but then you'll meet another error

Name "qs" already defined on line x

The solution

Python 3.10 has the cast function, and you wrap the above examples like this:

from typing import cast
# ...
qs = cast(MyModelQuerySet, self.model.stuff())
# ...
qs = cast(MyModelQuerySet, need_other_stuff.get_other_stuff(queryset=qs))

The cast function essentially just takes in a type as the first parameter, your value as the 2nd and simply returns the 2nd one unchanged.

# .../lib/python3.10/typing.py
def cast(typ, val):
    """Cast a value to a type.

    This returns the value unchanged.  To the type checker this
    signals that the return value has the designated type, but at
    runtime we intentionally don't check anything (we want this
    to be as fast as possible).
    """
    return val

A word of caution:

Although doing this is a way to inform mypy your intent it doesn't guarantee that the assigned value will be of the correct type at runtime.