I really enjoy how the author broke down both the problem and the solution, explaining why the existing solutions are inadequate. I often see articles about new products or new programming languages, and their features, but they often forget to mention what problem prompted the solution. Why was a new language invented to solve this problem? Why were the existing solutions limiting?
Bravo to the author here for making all that information clear and up-front.
> I really enjoy how the author broke down both the problem and the solution, explaining why the existing solutions are inadequate.
A thousand times this, I always get frustrated that every single product, tool or service has to have a lot of marketing gibberish, catchy slogans, etc., but zero explanations of what problems do they attempt to solve, or what was the motivation to create them.
You usually are required to dig further in the documentation or look for an external reference to understand what's the real purpose behind a product, tool or service.
I actually wrote that tool based on a huge frustration I had going back and forth to the browser, and not being able to work while being agnostic to the services, which felt being a totally broken UX. I have then worked on that tool for a while, used it and refactored it a lot up until I'm happy using it.
All in all, when I'm doing a tool, I'm first doing it for myself (I might invent the DIFY concept, to override the DIY concept )… only then I'm sharing it with others.
for a while now. It might be interesting to compare and contrast your tooling and git-hub - ingy's being straight bash, while a little crazy, makes installation really trivial, but then again yours supports multiple services, but then again who knows whether the different subcommand syntaxes will work better or worse for any particular other user.
(also, lest anybody think this is a veiled complaint, I'm sure ingy's reaction to finding your work exists will be "awesome, the more the merrier" followed by figuring out if you came up with any features he wants to steal ;)
Thank you, though pretty close, our approaches are slightly different. My main focus is use case, I did not implement the social features of a repository, because those are things that happen happily in a browser.
But project/merge requests/gists are things that have a lot of sense to happen in the commandline /instead/ of within the browser. So basically, his solution is more feature complete on the social side, whereas I implemented more the repository interaction side (like creating gists, or fetching a PR).
So for sure he's got a point for portability being a bash script, but still python3 is begining to be widespread enough.
I mean, if debian-stable got it, you can consider it's everywhere…
Not sure if it's something you'd call a social feature, but I find things like listrepos on an organization to be really helpful for piping into xargs to checkout everything a new customer depends on.
The PR stuff is so far as I can see complete in git-hub - what do you think you're missing?
(edit: if you meant gists, lots of the people I know already use paste site CLIs that handle that, btw, so to -us- that's more a social feature than the existing functionality is)
My bad, I did not scroll down to see how complete the command set was. At the top of the readme the list of commands is misleading (I thought there were only those at the first read).
I'm not sure I can see a real use for checking out every repo of an organisation (I mean, if you do that with mozilla, you better start before going for a long holiday ), but having it /possible/ is a great thing
So I guess we're pretty much alike, git-repo being strong for being service agnostic, and git-hub being more complete.
Though, in my redesign plan I got for version 2, I have in mind a framework to make it possible to implement service-specific CLI API, while sharing a same core CLI API for the service-agnostic features.
So my hope is that people will like the tool enough to participate in the development and help make the tool more feature complete across the services!
> I'm not sure I can see a real use for checking out every repo of an organisation
Maybe I wasn't explicit about it, but I did (and do) have one:
New customer with a github org and I want everything from their org so I can have a spelunk through the code while getting up to speed, build out dependency graphs to get a feel for how things interrelate, etc.
In fact, making it possible to script checking out an org's worth of perl repositories and shoving them into an http://p3rl.org/App::opan instance easily was the thing that was painful enough for me to finally install git-hub.
So, yeah, it may not be a common use case but it's actually an essential feature for me. Whether this means you should prioritise stealing it likely depends on how common it is, which I can't answer :)
I'm using git-hub as well and I find it very DWIM and more complete. (Have to add that I'm also a collaborator now, so maybe a bit biased)
I couldn't figure out how to use git-repo.
When I figured out which command I have to use to list pull requests for github, it asked me for my ssh key password. I was wondering why it would need my ssh key at all, so I tried -v, but -v did nothing.
Then I gave up.
Also, it lacks tab completion, as far as I can see, while git-hub supports it for bash and zsh. And since the order of the commandline arguments is not fixed (`request list` vs. `request <user>/<repo> list`) it might be actually difficult to add completion.
I'm developing a generic completion generator, which would allow to add completion to existing commands, but it's not able to deal with this dynamic commandline argument order.
I use hub[0] to accomplish some of the same things. `hub fork` creates a fork and sets up a new remote. After pushing to my fork, `hub pull-request` creates a pull request.
Well, I saw about hub, but I wasn't happy with the fact that it's a github only tool, I really wanted a service agnostic tool, and also because the CLI API was not great. I don't want a tool to replace git, I want it to extend it.
What git-repo really offers is a tool that integrates nicely to the git CLI (as `git-something` is also available as `git something`, so git-repo acts per default as `git repo hub…`). And then it autoconfigures the .gitconfig aliases to have the `git hub`, `git lab`, `git moo` you'd want.
So besides the fact that hub is certainly much more mature than my tool, it wouldn't be a tool I'd like to really use.
You can just alias hub to git to solve the first problem. As far as being service agnostic, that's a fair point. I rarely use anything outside of GitHub.
Hub literally crashed my computer, for some reason. I never worked out why, but it'd peg every core at 100% and macOS would crash! Weirdest bug, never worked out why that was the case, but I never managed to solve it. Bit hard to debug a system that's completely unresponsive!
I don't use hub much, but I prefer not to alias it even though that's what they recommend as well.
I think it's best to keep them separated so you know what stuff is coming from hub and what's in vanilla git instead. "git" is already synonymous enough with "GitHub" in many peoples' minds without including GitHub-specific functionality in your mental model of the command line client.
Well, TBH, as an engineer, I hate to reinvent the wheel. I prefer to join and improve the existing concept with my own ideas to make it the synthesis of many peoples ideas
You might see on my github that I do that a lot, I'm being some sort of mercenary of FLOSS projects, as I patch wherever I see a bug or want a new feature without really caring about the "big plan"
Actually, I wrote the tool because deep down I just want to /use/ the tool.
then you watch the code, create a branch:
git checkout -b `bugfix/nasty_one`
Please remove backticks from the command. What if you called the branch `poweroff` instead and a naive reader copy/pasted the command into her terminal?
Some recent distros will actually let you run that as yourself, without root, if you're physically present at the machine. It's convenient for personal machines.
See user chainsaw10. Anecdotally, on my vanilla Arch systems, I can run "poweroff" without being root, and without sudo, and the computers will end all processes and turn off.
Well, first the name really is `git-repo` not `repo`, and as nobody has really heard of it yet, the pypi package is already on first page of go, and second page of google.
I hope that if people like the tool, it's going to be easier to find.
Of course, I'm not really obsessed with the name (as anyway the tool is designed so the user does not type git-repo everyday ). So if someone comes up with a name that's great that can be better, I'm all ears!
About the fact that google has that other tool… Well they really are not much projects using it, beyond the google projects, and the tool really is called `repo`, not `git-repo`, so the name clash is not really there.
I really like that this works for all services interchangeably. Yes, Github created Hub for this same problem, but I've used Bitbucket at some jobs and Gitlab personally for some projects. Having something uniform and service agnostic is great. It's clear the creator thought a lot about this design feature.
This is wonderful. I'd guess that this little tiny bit of extra friction is enough of a disturbance to the workflow for private local/personal/work projects to prevent many from bothering to contribute back in some cases (I'm definitely guilty here).
And I hope also level the game for alternatives to github. Git is distributed by nature, and we should only be happy to see cross services forks happening!
note that I did not say that git has a good UX, I said it's /rather well designed/… Maybe should I have emphazised on the /rather/, or used something stronger like /relatively/ .
But anyway, designing the command line API of my own tool took me enough headaches — and I'm still not 100% satisfied — to understand how hard it is to have a consistent and great UX for something that offer as many features as git does. Though, I'm trying my best to avoid the decade old troll about git's UX (like vs mercurial)…
In the end, what I try to make clear it's that I'm not trying to replace what git does, but cluster a bunch of repository actions tied with some service's API operations.
And I try my best to offer a pleasant UX doing so. But please, help me making it better!
I agree with you on everything, except the "easy to remember" part. Git's UI is very convoluted and inconsistent. Designing a good UI is really hard, but I see too many people fall into the "I learnt it, so it must be easy!" trap.
Congratulations for making git-repo, by the way! It looks extremely useful.
The name is slightly problematic as Google also released something with the same name: used for Android to solve the issue with multiple repositories instead of submodules.
Besides that, looks interesting and I will give it a try!
Well as I said for another post I'm aware of that. As the google project is mostly used only for android and chromium, and as the canonical name of the tool really is /repo/ and not /git-repo/, there's no real name clashing.
The only issue might be my project being shadowed by google and go, but the more people will like and use the tool I baked with , the more it'll appear on their first page. So it's not really up to me, but up to you all liking it!
And finally, if someone comes with a nicer better name than what I chose (I actually did not spend more than 2.42 minutes on the name when I had to mkdir the project's directory), I might be happy to rename :)
Otherwise, I hope you'll like the tool and send me patches and <3
Goood, I'd really like to try it out. I'm tired of opening the browser go to my fork and create a branch to make a MR/PR for the project I'm working out.
Bitbucket support is there. But it's not feature complete, and not up to my standards…
The issue I have is that I got to implement OAuth2 for bitbucket, because the privatekey scheme offered by bitbucket does not work for a CLI tool use case. So currently, the credentials (login/password) of bitbucket are stored in CLEAR in the configuration file, which really, really sucks…
And some new features (like the snippets) are not implemented in third part python libraries, because most of the libs are old and unmaintained.
Anyway, following some discussions I had with atlassian bitbucket developers, bitbucket (like gitlab) is planning to switch to swagger, which will help automagical generation of client-side feature complete libraries. So I'm planning to level bitbucket feature set as soon as they land the new API.
So I hope it won't be long before all the features will land for the three services!
I'm amazed that nobody has commented yet about how this is a CLI wrapper around web services which are wrappers around a CLI tool. Or is everyone here already too jaded and so used to piling endless abstractions upon abstractions that this is considered normal?
How about if people simply accepted patches via e-mail, instead of requiring those obnoxious github pull requests?
See the issues, it's already implemented, but not released yet. The issue is that I also am working on the gitlab snippets… which lacks an API. So I'm currently working on a patch to add an API to the gitlab snippets, so my tool can have gitlab snippets!
Honestly, I just opened the post to find ridiculous things to criticise. But in fact this looks like a valid helper. Sure I don't have the same workflow, but this one seems to work as well. Good work!
I'm surprised Github et al. haven't tried to contribute some functionality for doing this to git. The thing described in the article is extremely common.
Bravo to the author here for making all that information clear and up-front.