Build AWESOME CLIs With Click in Python

[@ArjanCodes]

💡 Get my FREE 7-step guide to help you consistently design great software: arjancodes.com/designguide.

[@jurgenrusch4041]

Hi Arjan. Thanks for yet another super clear and instructive video. And,... at the right time as I am creating a CLI for various related functionalities already written in Python. So, again thanks! Un your video I saw you were using poetry. As this was not the main focus of this video you simply executed poetry commands. I looked back in your video history and saw that in 2023 you posted a video titled "How to Build a Complete Python Package Step-by-Step". There you used setuptools for packaging, and not (yet) poetry. Even after many years of Python programming I find packaging a complex topic. For my work I would very much like to use poetry for packaging and distribution. I would really love to see you (and your team) create an in-depth video about poetry, covering topics like a build frontend and backend.

[@EusebioResende]

Click is a great library. A couple of features I used were Multi Command Chaining and Multi Command Pipelines. May I suggest a video with some advanced features of Click. It would be awesome. Great video Arjan and thanks for sharing.

[@maleldil1]

I don't really see a point in using Click directly anymore. Typer does the same thing but better, using type hints to set up everything. It's also based on Click, so if you need something that Typer doesn't offer natively, you can use Click through Typer to do what you want. IMO, Click by itself doesn't offer enough to warrant adding a dependency, as stdlib's argparse is quite powerful and easy to use.

[@traal]

Agreed, I also use Typer for everything. And I agree that argparse is pretty great. In fact, unless I really need subcommands and composability, I personally prefer argparse over Click. There’s also a nice project on GitHub callled “argh” that wraps argparse with type hints, very much like Typer, but much more lightweight.

[@SampadMohanty7]

Or Fire by google.

[@LucidProgramming]

For that matter, even justifying a third-party module like Typer might be overkill given what you can achieve with just argparse as well. I find that to be completely adequate in most scenarios and does not bloat the dependency tree of the project.

[@maleldil1]

@@LucidProgramming argparse is fine, but I've found to be much better, especially with handling subcommand and drastically decreasing boilerplate.

[@DJStompZone]

> Also sometimes called a flag Correct me if I'm wrong, but my understanding is that flags are boolean values (they're either present or they aren't) and don't require any additional information, whereas options are necessarily followed by a value. Sometimes they can go either way, depending on context. For example, "--help" might be a flag if used by itself, or it can be an option if used like "--help ".

[@ArjanCodes]

That's right, a flag is another word for a boolean option - that wasn't clear in the video.

[@DJStompZone]

Right on, cheers 🙌

[@Aceptron]

Nice video!! I would love to see a video on most commonly used libraries for creating cli programs in python in the industry with their pros and cons

[@ladycoder2095]

This is absolutely fascinating! IAbout nine years back, had used click to create a command line tool which took a file list of tsv and output a combined xls file with the original tsv files as sheets 😀. This has been a nice refresher.

[@bryangarcia9488]

Great video. I tend to use classic argparse at my workplace but I’ll give click a try for future projects 👍

[@iamrodos]

Yes to pytest. As I was watching I was thinking this video would be even better if it use TDD right from the beginning where pytest and coverage was used. Create the empty functions, write the test of expected behaviour, complete the functions, write next set of functionality, rinse and repeat. Would demonstrate a great way to "approach" development that would help a lot of people see some good technique and practices.

[@iamrodos]

Great video BTW.

[@YazzDAtlas]

I would like to see some practical test.

[@ArjanCodes]

Thanks! Though I agree that it is good practice to follow a TDD approach, I’m afraid that it will slow the videos down too much. But I might test it in an upcoming video 😎.

[@Deiha-z4d]

How to test CLI would undoubtedly be a great addition to a possible continuation of this topic.

[@Geo-Artist]

Great job! Please prepare more extended example with pytest, pydantic It's a pleasure to learn programming with you! Greetings!

[@patrickkurmann]

I liked your introduction story. It instantely got me hooked

[@riki4878]

Great video Arjan :) Good job. We use click in our company everywhere. It's really nice package.

[@ArjanCodes]

Glad you liked it!

[@snakehunt89]

I would love to see an additional video covering some more advanced usage like adding a logger for scripts or conditional parameters (like if one option/arg is supplied make another not required)

[@MohsenPadidar]

Thank you for the professional content you produce. I learned a lot from you. ❤ Please do more rust

[@ArjanCodes]

More to come!

[@realpdm]

Thanks for sharing this. I've been using docopt for a long time and while I like it is can be very finicky to use. Click looks a lot more deterministic without all the hassle of using argparse directly. Will definitely try it with my next project.

[@ArjanCodes]

You're welcome - glad you liked it!

[@phortheman]

Looks cool! I usually write my more complex CLI tools in Go with Cobra which has a great interface. This looks kind of similar but for Python instead!

[@mmilerngruppe]

21:21 environment variables sets are stored in terminal history too if they are not coming as exports from other scripts. but if they come from other scripts, they are already somewhere stored. it's a uroborus problem. may be I am complete wrong about it, I would like your opinions on it.

[@Deiha-z4d]

21:40 Perhaps the 'platformdirs' library is a good option for determining the location of configuration files.

[@johnsontea8542]

Yes would love to see how to add pydantic to add validation to configuration values and with a logger and pytest!

[@j4g094]

I am a DevOps Engineer. Any time I need a CLI tool to help with some task or another, I write it in Python. That is because I have yet to find anything I needed to do on my Linux servers, for which a package did not already exist. For almost all of them I use Click and Rich and when I don't, it is because I need a very quick solution without dependencies outside of the standard library.

[@mmilerngruppe]

3:41 Arjan, what text editor do you use?

[@iaroslavdavydiak6439]

Thanks for interesting and detailed explanation!

[@taufiqurrahmansagafkelrey2833]

favorit coding uncle, finally!!!

[@DanielRodriguez-lu3uu]

Great Tool, quick question: something that grabbed my attention was the part where you installed "notes" so you don't have to call python on shell. Is that only possible because you are using Poetry? is there another way to achieve the same result with out Poetry?

[@alexandrodisla6285]

Click, Typer and Rich

[@express4863]

Click and Rich are my two "go to tools" for anything CLI related. Using things like rich.Live() for updates for long running or slow application updates really improves the user experience.

[@pranaypallavtripathi2460]

are typer and rich also python packages for building cli tools?

[@express4863]

@@pranaypallavtripathi2460 typer has some overlap with click. Rich is more for output to the console. Rich makes working with the console in an interactive way much easier. I use it even when I am not making a CLI application.

[@PhunkyBob]

As always, thanks for sharing. ❤

[@ArjanCodes]

Thanks for watching!

[@spanomatic]

Thank you Iron Programmer!

[@ArjanCodes]

You’re welcome- glad you enjoyed it!

[@klmcwhirter]

I looked at click sometime ago myself, but am now just using docopt (NOT docopts - with an 's'; which is for shell scripted CLI tools). docopt allows you to specify the CLI in the main modules docstring. You don't have to use that convention, but it is the most practical. The CLI documentation IS THE parser spec. Add one line of code to get the parsed CLI verbs, options, flags, etc. Click is intriguing, but there is too much magic and the interface still needs to be documented - which never happens in practice on a busy team.

[@Julie9009]

+1 for docopt

[@DiegoRoccia]

click and rich together really help creating great CLI interfaces in python. too bad distributing python code to end users is so complicated. For this reason we recently migrated one of our internal CLIs to golang

[@adrianabreu1565]

Would love to see the new pydantic version implement in the todo app!

[@helboy1111]

I used to love click, but since finding out about invoke I don’t think I’ll ever create a CLI app manually anymore ^^

[@Deiha-z4d]

20:15 seeing pydantic in this context would be very interesting.

[@doodah120]

I like to have my config object loaded in the top level __init__.py, that way it is globally available and doesn't need to be passed around. Is there any issue with that?

[@the.elven.archer]

I would love a guide like this but for ncurses-like applications :D

[@shridharcs]

How about permission handling with click?

[@ShizoMoses]

I just noticed that Arjan has "notes-py3.12" instead of "venv" - that is so much more useful to see, how do I do that? Does anybody know? :)

[@MyrLin8]

LOL ok got your email, "helping students get drunk :)" hehehehe, ok not really, but the fridge issue is closely related yes? Now we'll learn how to write a CLI, since I'm going to need one soon. Bookmarked!

[@ArjanCodes]

I remember when I was a student, I didn’t need all that much help to get drunk, haha.

[@nickeldan]

I like your keyboard! What is it?

[@ArjanCodes]

I think I’m using the Nuphy Air 75 here.

[@FragrantVagrant69]

I have a question, hopefully you can help. I wrote a basic gui app for a friend to help them solve a problem at work. They showed their boss and he loves i and wants to distribute it to hundreds of users across the country because it will save them hours of work every day. I packaged it using pyinstaller and windows blocks it unless you disable windows defender. They dont seem to have the option to exclude the file from defender, so i wanted to ask what would be the best way to distribute the get this app recognised that it isn't malicious? I heard that signing it doesn't guarantee that it will be trusted and its such a simple app i dont know if if its worth submitting it for review or something like that. I don't really want to rewrite it as a web app as a simple exe is sufficient for what it does . Do you have any suggestions?

[@DJStompZone]

My advice is not to use pyinstaller. Defender is gonna freak out every time, unless your executable is signed by a trusted certificate authority. If you want to proceed, you might consider getting your application officially signed. Otherwise, you'll need to instruct users to install and trust your certificate manually through the certificate manager in Windows, which is inadvisable unless you know exactly what you're doing.

[@simonchpmnk4392]

@9:43 kzbin.info/www/bejne/fIjElJSkqNGZfK8 Line 14: TypeError: unsupported operand type(s) for /: 'str' and 'str' Need to use e.g. pathlibs Path or similar.

[@ramukamath2110]

i dont think you need to run pip install -e . . poetry install will install current project in editable mode. you just need to activate the venv using poetry shell

[@AndyWallWasWeak]

9:35 now we are cooking!

[@ganeshprasadrao5596]

typer ?

[@duncangibson6277]

Interesting video, but... New users are probably at Python-101 level, If they read the docs, they will find command line options parsing using getopts, optparse, getopts, docopts, and maybe others. It would be interesting so see a video that compares and contrasts these different modules. Secondly, new users are probably not 100% conversant with @decorators, especially the multi-level decorators shown in this video, so it would also be interesting to see the basics of how users work with decorators, and maybe more importantly, how these modules create these decorators, Apart from that, keep up the good work 🙂

[@drac8854]

How about you remake this software in clap(rust) 🙂

[@בןרונן-ח1ש]

Call it "Arjanotes"

[@ArjanCodes]

I like it. 😁

[@Amulya7]

I have heard this story before. I don't remember the title of the video.

[@siestoelemento1019]

Ok

[@moshedicker6786]

Typer

[@rdmferreira]

ok.

[@noobymemer]

mynotes😂

[@Prevok]

Still would rather use argparse. I am forced to support projects using click at work, but I so don't see the point. Rather than have everything define in one place, it's all over the place with tons of decorators. It might be "easier" to create positional arguments, but at what cost...

[@robotnaoborot]

click is still painful to use., i prefer google's "fire"

[@johnathanrhoades7751]

I love dotfiles in the $HOME/.config/ directory. Just makes it really easy to keep everything together. That or a dot file directly in the $HOME directory.

[@HarshVerma-k9z]

Cyclopts library is much better

[@jylpah]

Cyclopts looks really nice! I went from ArgParse to Click to Typer. support for Literal, Union and asyncio is great. I will try this next time

[@Schlynn]

Car to give even a single sentence for why?

[@HarshVerma-k9z]

​@@Schlynn discord.com/channels/820628246965780520/1186057034873962627 You can check this thread on discord, I had a long discussion with the owner of cyclopts. It will explain why it's better than other cli parsers. I've tried Argparse -> Click -> Typer -> Cyclopts. Cyclopts is the easiest to work with (has some really good features) and it's carefully maintained. This was the first reply on posted on the thread: I agree that typer is bloated. For example in Enum type, the keys should be the user input but in typer library it uses Enum values as user input. I planned on creating a PR, but I don’t like its codebase at all. Cyclopts look interesting btw,

[@cameronball3998]

haven’t had a chance to watch the video yet (on my watch later lol), but just want to throw docopt’s name here. i use docopt on all my projects now as it’s absurdly simple and my documentation becomes code. maybe for a production grade or release application i’d go with something more heavyweight, but i find for the vast majority of use cases, docopt is clean, easy, and reliable, and has the features i need (e.g., nested mutually exclusive argument groups).

[@nahiandev]

I'd rather do it using C++.

[@hriscuvalerica4814]

Real programmers use real programming languages like Fortran not toy language like C++

[@PhunkyBob]

@@hriscuvalerica4814 Pfff... There is nothing better than assembly language. 😆 How much time would last a YT video of a tutorial about making a CLI in assembly to create a notes application 🤔

[@nahiandev]

@@hriscuvalerica4814 I request you to not consider me as a 'Real Programmer' then. Peace 🙏

[@ssmith5048]

only a sadist would create a cli tool with python - or they would build it and install it on their worst enemies system.....

[@ArjanCodes]

A true sadist would then post a video about that on KZbin, which would lead to even more CLI tools built with Python.

[@ssmith5048]

@@ArjanCodes go for it if you don`t care about runtime

[@keeganlaporte4758]

I maintain a Python CLI tool. It is much faster. Speed, three kinds: speed of program execution, speed to develop, speed (time) to maintain issues with the program. My Python CLI is much faster, saving weeks in the second and third category. It’s unlikely that the first category will ever compensate for time lost in the latter two. Python is typically the faster language when you consider the real problems. If you really convinced your customer to pay the costs of premature optimization, go ahead and use cpp I guess. I couldn’t imagine trying to convince our customers to wait extra weeks for bug fixes and development

[@therealslimaddy]

@@ssmith5048cli tools are very much viable in python. If it’s slow then there is a skill issue clearly.

[@ssmith5048]

@@therealslimaddy i would call it a poor decision issue to choose python for a cli tool for anything not trivial, instead write tools in c or cpp for cli tools. python for the quick and dirty scripting it is tolerably good at.