I know I don't know git, that's why whenever I have to do something new I research it for 30 minutes and then make a script so I don't have to do that again. Mad respect to all y'all who somehow find the time to master git and all the other devops wizardry on top of learning your language of choice.
I think for a lot of people it's less about time and more about version control being this necessary evil we have to deal with when really we just want to write code. I'll spend hours learning programming languages that I'm never going to actually use instead of spending that time reading a manual on the version control that I use every day. I'm currently trying to convince myself that it is one of those things that are more fun once you actually know it.
@@MHLuttekesI struggle with this too, my guess is because these style of presentation just keeps bombarding you with information so you’re more inclined to stay focused. Whereas usual presentation are sprinkled with useless fluff that makes you “turn off” your brain for a while and “turn on” again when the important bits were presented
It's better to watch it later (vs. live). I have to pause the video, take notes, and think about how it would apply to my job if I want the info to stick. This is a great presentation, though. There is zero fat here.
Timestamps for dorks: 00:00 - Introduction 01:06 - About Me (well, Scott Chacon) 02:36 - How Well Do You Know Git? 05:09 - Our Agenda 06:25 - Some Helpful Config Stuff 09:42 - Oldies But Goodies 16:22 - Some New Stuff (You May Not Have Noticed) 23:48 - Some Big Repo Stuff / Monorepo Stuff 33:29 - Some New Github Stuff 35:54 - GitButler 36:50 - End of talk 37:03 - Start of Q&A Session 37:06 - Q: Why does GitHub not do git range diff? 38:28 - Q: Why do submodules suck everywhere? 40:16 - Q: With SSH signing, is it possible to specify more than one key? 40:42 - Q: Why can't --force-with-lease be the default force? 42:33 - Q: If you were back on the Git development team, what direction would you like to see it move in? 44:58 - Q: We all love the Git CLI - but do you ever use any visual tools? 46:41 - That's all folks!-
`git blame -C -C -C` is pure gold. But wait, there's more! From `GIT-BLAME(1)` -C[] In addition to -M, detect lines moved or copied from other files that were modified in the same commit. This is useful when you reorganize your program and move code around across files. When this option is given twice, the command additionally looks for copies from other files in the commit that creates the file. When this option is given three times, the command additionally looks for copies from other files in any commit. is optional but it is the lower bound on the number of alphanumeric characters that Git must detect as moving/copying between files for it to associate those lines with the parent commit. And the default value is 40. If there are more than one -C options given, the argument of the last -C will take effect.
So excited to share this with my boss Mike, he's our local Git wizard and I'm so thankful for him. He's always so excited to share the new things he's learned and I'm looking forward to paying it forward!
Thanks for the book. It saved my neck in 2011. Every intern I've had since 2012, on their first day, I give them the book and ask them to read it all, and do all the exercises. I give them paper so they can mark it up as needed. Every intern that does this, has no issues with git all summer. The one or two interns that did not do the exercises? They struggle. Great book!
Pro Git 2nd ed. Edition by Scott Chacon -- just do every exercise and try every command, even if it's boring or you have done it before. That's what we learned over 10y of interns.
lmfao, you ask every intern to read a 500 page book for a summer internship? try not to respond with: "yeah I work at a super awesome place and im so smart, try putting a little more work in kiddo, pull up your bootstraps etc etc"
@@ronminnichhope you at least pay them for that lmao seems like a shitty workplace ngl. I have never had any problems with git and I didn’t get forced to read a 500 page book. If I wonder something I’ll just google it. Takes 3s
Scott Chacon is a legend! I learned git in 2016, and I love it, it's one of the most important tools for me, after start my journey as a developer, I start to teach git, and the best resource always was the Pro Git book.
committing to multiple branches is brilliant! its always a PITA to notice/fix a core bug when working on a feature branch.. stash->checkout->pop->commit->checkout->merge is so much overhead for what started as a 5-second one liner fix.
This talk is great it felt I knew nothing of Github. But it was super fast delivery before my mind , eyes, and ears can align together and comprehend anything out of it
Best use for GUI Git has to be staging and unstaging parts of a file. In VS Code, just select some lines in the diff view and right-click. Stage selection and unstage selection. Amazing! 😊
Virtual branches are an amazing feature. There are so many times I've been working on a branch and seen a minor fix in a file unrelated to my work. Streamlining those fixes while keeping them separate from my branches is very efficient for everyone.
46:30 I've been using Git for a long time and I use git gui for building my commit because I always commit lines of code, not full files. And I also use gitk for reviewing history of the code unless I'm looking for some highly specific thing that gitk doesn't support well - for those cases it's mostly `git log -p - - optional/path/limit` and grepping the results if needed. (The dashes should obviously be combined without a space but UA-cam comment interface cannot display that string correctly.)
@@jasonhenson7948 They kicked everyone out who did not have a seat before the talk started. We were sad, but the fire warden would have remained happy.
I was kinda hoping he would talk about git worktrees, I use those to checkout branches into separate file paths on my computer which allows me to work in different branches simultaneously. And I can keep long-lived branches separate. This also integrates really well with vscode workspaces.
frankly i have an old habit of sidestepping Git completely for that, i simply checkout the repo into two independent copies and i switch between those : )
I loved your "Ask your code" talk back at Code.Talks and I love this one even more! Holy shit that git butler idea is great, I've always wanted to work on several branches intermittently
Remember when version control used to be simple? When it didn't need to be wrapped in huge frameworks so that people could work with it? When it didn't need conferences like this? When it wasn't the main topic of conversation on a project, and we just did the project?
Yes I do. And source control back then wasn't a tool the way git is. Like, also,git is awful in many ways, but I still love it deeply for the things it lets me do, manipulate and interrogate my repo history in many useful ways. It lets me approach certain problem classes in a meta way that old source control just could not. Branching is a first class citizen in git and that was not the case inb4. Learning git is a brick wall, as an early adopter I had the difficulty of explaining git many times to many companies, simple things are often stupid, and git basically requires you to learn it's mental model pretty well before you can do basic things with confidence. Then again, idk, sql doesn't make much sense unless you know what a relational database is. So git is hard in dumb ways, and if anything ever replaces it the way git ate svn, then I can't wait to drop git over whatever that future better system is, but it doesn't seem to exist, so I'm on the whole far far happier with git, warts and all, then any prior system. CVS wasn't even atomic, and vss would corrupt itself, svn really couldn't do branches and merge conflicts ruined your working copy in unrecoverable ways, and git is so so fast, and also best case was you could do exactly one local commit effectively where git you can do infinity local commits and also peer to peer stuff. Evolution of source control systems has been a net positive
@@rebeccakeller4666 Hi Rebecca. I could write a similar passage about how bad it is, but I won't bother. I will just say one thing; it is too difficult, and any good software should be easy. Yes, it can be learned, and when you have learned it, it is ok. But it shouldn't take so long to learn. I have been using Git for the last 10 years, so I feel like know it fairly in depth, but it took me about two years to get to this level. Why? You can of course take the easy way out at this point and just say I must be stupid, but I will assume you won't do that. It took me just a few weeks to get to a similar level with CVS, SVN, Clearcase etc. Now I know Git, but I have to work on projects with more junior people who are currently going through the same hell I went through. While they are learning, they make a lot of mistakes and people have to spend their time trying to fix them. The old systems, which would have improved a lot by now, were much better because they were simple and just worked. Anyway, I am glad you like it. Each to their own.
@@vibovitold Creating branches was just as easy as it is with Git. But you didn't need to make a new branch for everything. If it was a simple fix, you just checked out the code, did the fix, and checked it back in again, via a review if needed. I know I'm flogging a dead horse here, but Git is a huge step backwards.
@@jimmorrison2657 creating, kind of. merging, now that's not how i remember it at all.: ) and i'm not a fanboy of Git. it is poorly designed especially in terms of API - commands are poorly named etc. but this actually emphasizes the question - if it beated all the competition DESPITE those blatant flaws... why did it happen in the first place? what's your theory
I still use RCS for personal projects involving a single file. For multi-file projects that I'm not making available publicly I'm more likely to use SVN . I started using git only because I was working on some Open Source Projects that used it. I don't like git that much. It is the only VCS that has resulted in me losing some work I had done but not commmited.
Great talk, I really enjoyed. I was wondering if the virtual branch is something different to what you can achieve with `git worktree add ` ? Amazing, helpful talk. Thanks
It is for the same problem, but an important difference. Worktrees give you a different working directory per branch, so if you have a feature and a bugfix, you don't have anywhere where you have _both_ the feature and the bugfix in one direcotry at the same time. With GitButler you do them both in the same directory and make branches from the hunks as though you were `git add -i` each hunk and committing. In other words, we keep the work of the two branches separated in memory rather than on disk. This is very nice if you want to see how branches will work together before they're technically merged in git.
"Unable" to use major tools??? Wasting time to use git blame in cmd line to look cool while it's a click away is being Unable to learn everyday tool? I can't live without cmd myself but this is not the type of stuff one use cmd line to make their life easier....
@@MJ-xg2ow if it's available in your editor, it's fine. Some features are indeed more useful when visualized. But this doesn't mean that knowing fundamentals is useless.
This is the first time I heard of porcelain commands. A quick google and... it makes sense! Part of most software is plumbing that has to work, but is ugly, and not everyday user-facing code. The nice user-interfacing stuff is porcelain.
43:10 - Working on multiple branches at the same time / multiple indexes, heads: Isn't that what `git worktree` does? I use it quite a bit, especially if I need to make a change to one branch while I have a bunch of other uncommitted changes in my main working tree. I also use submodules - I have a lot of 3D printing projects using OpenSCAD, and I've developed a common library of functions / transformations / etc. Since OpenSCAD doesn't have a package management / dependency system like Python or NPM, I just add my common library as a submodule to each individual project.
My friend's company is switching from SVN to Git after decades and the one piece of advice I told them was the reflog will save you from almost anything!
🎯 Key Takeaways for quick navigation: 🛠️ Git configuration: Learn to set up aliases using `alias` and `git config` for better workflow customization. 🕵️ Git Blame: Utilize `git blame` with options like `-L` and `-w` to get detailed information about changes in specific lines and ignore whitespace. 🔄 Git Log Options: Explore options like `-S` for pickaxe search and `--word-diff` for word-based diffs in `git log`. 📜 Git Reflog: Leverage `git reflog` to view a log of reference updates, helpful for tracking changes, especially after operations like reset. 🔄 Reuse Recorded Resolution (`rerere`): Enable and use `rerere` to automatically handle recurring merge conflicts, saving time during rebases and cherry-picking. 🌐 New Git Branch Options: Discover the `--column` and `branch.sort` options for `git branch` to enhance the display of branch information. 🌟 Git Column Command: Learn about the new `git column` command, a simple tool to format input into columns. 🚫 Force with Lease: Familiarize yourself with the `force-with-lease` option, a safer alternative to force pushes, preventing accidental overwrites of others' changes. 🔄 Force with Lease: A safer option for force pushing in Git. It checks for stale references before pushing, making it useful for scenarios like rebasing or amending commits. 🔐 Signing Commits with SSH: Git now supports signing commits using SSH instead of GPG. This is beneficial for those facing issues with GPG and provides a similar level of security. 🛠️ Git Maintenance: Running `Git maintenance start` in your repository adds a cron job for background maintenance tasks, improving overall performance by optimizing tasks like commit graph generation and loose object collection. 📂 Large Repository Handling: Tips for handling enormous repositories, with insights from Microsoft's efforts in making Git work well for large projects like Windows. This includes features like prefetching, commit graph, file system monitoring, and partial cloning. 🏗️ Sparse Checkout: Useful for monorepos, sparse checkout allows you to clone specific directories, reducing the amount of data fetched and improving performance. 🔄 GitHub Merge Options: GitHub offers various merge types like merge commits, rebasing, and squash merging. You can enforce these standards and auto-merge pull requests based on CI status. 📝 Linear History Requirement on GitHub: GitHub allows you to enforce a linear history, rejecting pushes with merge commits. This is useful for maintaining a clean and straightforward project history. 🔄 Pull Requests with Heads: Pull requests on GitHub have corresponding refs, enabling you to pull down and merge changes directly, providing an alternative to using the GitHub interface for merging. 🔄 GitHub allows fetching from various refs directly, not just adding remotes for pull requests. 🌿 GitButler introduces virtual branches, enabling simultaneous work on multiple branches without switching. 🚀 Virtual branches in GitButler facilitate handling features and bug fixes separately without commits into the feature branch. ❓ A question is raised about Git range diff, with the speaker suggesting it's not widely used in typical workflows. 🔄 Submodules are criticized for their limitations, and the trend is shifting towards mono repos for better project management. 🛠️ Suggestions for Git improvement include the ability to work on multiple branches simultaneously and automatic recording of all changes. 🚀 The speaker discusses their preferences for using both command line and GUIs, depending on the task, with command line being faster for most operations. 🤔 Future directions for Git are explored, with emphasis on improving the user experience for modern workflows, such as multi-branch development and comprehensive change tracking.
Working on multiple branches can be achieved to a certain extent by using worktree command. It will give you the ability to have multiple worktrees pointing to different branches، and the ability to switch between them while keeping the changes. It isn't the easiest to use, but better then nothing.
The recording feature would be great. We've all been there, when we did a lot of work and then accidentally deleted the code, and it wasn't added/committed yet. And you pull your hair out wonder why th you didn't commit it... and you have to rewrite everything. It would be great if git allowed you to undelete something.
4:02 ahhh man, the plumbing commands was what I was REALLY hoping to hear more about.. I often like to temp-merge a big selection of branches together, write that to a tree, then commit the tree with a massive bunch of parents, of where I merged from.. Give myself a giant octopus merge and then throw away all the temp-merging commits, so you have a nice clean "MERGE ALL THE FEATURES AT ONCE" commit.. Looks fantastic :D
Hehe. Starting to watch the video. I am using git for 15 years (and before that hg, perforce, svn, cvs), and I would say I do not know git even on the surface. I know more about git in my company than anybody else probably, but I still would say I do not know git. And I do not want to know it.
Same here plus sourcesafe, clearcase and tfs. Used hg for many years, wrote many extensions for it, even up-streamed some fixes in the early days when windows support was shaky. Git is the best by far.
The fact that a nearly hour long video like this about Git even exists really shows why Git needs to be replaced by something better, something intuitive. Start by not naming commands that do different things with synonyms lol. Git is the JS of source control systems.
The naming is horrible, but the list of Git wrappers trying to provide a more friendly facade over its API longer than my arm, and yet none of them have ever gained much traction. Of course it doesn't mean they aren't legitimate improvements. I would liken it to the QWERTY layout. It's demonstrably not an optimal layout, quite poor in fact, but inertia is a tremendous force, so despite -perhaps precisely because - the number of alternative layouts is huge, none of them has really caught on.
@@vibovitold Yeah but you have to factor in the pain level and/or benefit in alternatives. Is there really a significant benefit to a different keyboard layout? Most people don’t even touch type much less do it all day long without lots of pauses. So gains for most is zero to close to zero. I think the Git tools trying to make it intuitive are pretty common. Every IDE of note has at least one, then there’s a lot of stand alone ones. Most devs aren’t using Git on the command line for sure. But even with these tools you’re still dealing with an over complicated change control system that just wasn’t designed for internal teams but for loosely collaborating groups working on OSS. For doing day job work at companies and gov orgs, it’s just not the right tool for the job but we got stuck with it till some new trendy tool comes out.
@@vibovitold That's not a fair comparison. There are very few people who are dissatisfied enough with QWERTY to push for a replacement. Most people seem so neutral on the topic it'll probably never cross their minds that an alternative could exist. The dissatisfaction with Git is much more tangible: I've never even used another version control tool and I still feel something better than Git should exist.
they're fine, but they're not together. "at the same time" meaning with both branches applied to the same directory. like being merged but still being able to work on them independently. this is what we do with GitButler with virtual branches.
Had I beenin the audience, I would have asked this: I am a rebase guy, I organically hate merging master/main into my feature branch. This strategy works great as long as I work on a solo project. The moment a conversation starts and comments/requests to your changes land in, the problem kicks in. When you rebase to refresh your master/main, then git push force, the comments usually get misplaced, the refs shift, and, most of all, your reviewers lose track of what has been changed since their last visit. Is there a golden solution to that?
Nope. Unfortunately the best way to preserve history is by not rewriting it. You could squash merge so the whole branch becomes a single merge and try to incorporate comments as commit data, but otherwise it's quite difficult unless you just merge.
@@gitbutlerapp Thanks for your reply. I'm at least glad I'm not overlooking anything and it's just the way it is. I hoped though, there was a piece of AI that was able to bind the overwritten history with comments ;-)
Rebasing is better for going main->feature, and merge is better for going feature->main. The more authors are contributing to a branch, the less you want to rebase it, and the better an explicit merge commit is for tracking. IMO.
@@LA-MJ Here is the problem. If someone during PR asks you for changes then you cannot really git rebase -i anymore and group the changes with the commits where they belong. So no, the problem is not solved, your history is only neat until the original commits have been pushed, all that follows is still a mess.
I assume why the filter for clone takes "blob:none" for no blobs but "tree:0" for trees is because the first means "no blobs", but the other doesn't mean "no trees" but "root of tree", i.e. index 0.
I know I don't know git, that's why whenever I have to do something new I research it for 30 minutes and then make a script so I don't have to do that again. Mad respect to all y'all who somehow find the time to master git and all the other devops wizardry on top of learning your language of choice.
If someone has the confidence to say they know git, I'd be scared of them. They're too powerful
I think for a lot of people it's less about time and more about version control being this necessary evil we have to deal with when really we just want to write code.
I'll spend hours learning programming languages that I'm never going to actually use instead of spending that time reading a manual on the version control that I use every day.
I'm currently trying to convince myself that it is one of those things that are more fun once you actually know it.
@@w花b or lying to themselves.
Try out GitButler, maybe it's a nice compromise. :)
I have found my people @@mr.mister311
The excitement level of this guy is priceless. Imagine working with him in a team!
Great talk, I really appreciate the "shotgun" style presentation - it cuts the crap while keeping it clear.
I don't know if it's ADHD, but my brain goes so well on this type of content where normally I struggle to follow presentations.
@@MHLuttekesI struggle with this too, my guess is because these style of presentation just keeps bombarding you with information so you’re more inclined to stay focused.
Whereas usual presentation are sprinkled with useless fluff that makes you “turn off” your brain for a while and “turn on” again when the important bits were presented
I also like he goes directly to the point instead of talking during 10 minutes with an history about versioning systems...
It's better to watch it later (vs. live). I have to pause the video, take notes, and think about how it would apply to my job if I want the info to stick.
This is a great presentation, though. There is zero fat here.
Timestamps for dorks:
00:00 - Introduction
01:06 - About Me (well, Scott Chacon)
02:36 - How Well Do You Know Git?
05:09 - Our Agenda
06:25 - Some Helpful Config Stuff
09:42 - Oldies But Goodies
16:22 - Some New Stuff (You May Not Have Noticed)
23:48 - Some Big Repo Stuff / Monorepo Stuff
33:29 - Some New Github Stuff
35:54 - GitButler
36:50 - End of talk
37:03 - Start of Q&A Session
37:06 - Q: Why does GitHub not do git range diff?
38:28 - Q: Why do submodules suck everywhere?
40:16 - Q: With SSH signing, is it possible to specify more than one key?
40:42 - Q: Why can't --force-with-lease be the default force?
42:33 - Q: If you were back on the Git development team, what direction would you like to see it move in?
44:58 - Q: We all love the Git CLI - but do you ever use any visual tools?
46:41 - That's all folks!-
Awesome! We're stealing this and updating the description. Thanks for the hard work!
@@gitbutlerappthe sarcasm lmao
You are real MVP
@@gitbutlerapphow to save work :) xD
can you pin this comment? @@gitbutlerapp
The talk is great, but the manner of it being presented is even better!
Suggestion: add chapters (timestamps) to this video. It really helps when you're watching for the second time!
If someone adds the chapters as a comment, I think youtube might catch them automagically
ask and ye shall receive. timestamps up above (or below... or wherever my comment ended up floating to.)
@@nlovelldev Already your FAN brother or sister or whoever you are. 😊♥
@@nlovelldev tysm
x
or the third ✋
Next to being an informative talk, the way it's presented is great. Good job on that presentation skills!
10 out of 10
`git blame -C -C -C` is pure gold. But wait, there's more!
From `GIT-BLAME(1)`
-C[]
In addition to -M, detect lines moved or copied from other files that were modified in the same commit. This is useful when you reorganize your program and move code around across files. When this option is given twice, the command additionally looks for copies from other files in the commit that creates the file. When this option is given three times, the command additionally looks for copies from other files in any commit.
is optional but it is the lower bound on the number of alphanumeric characters that Git must detect as moving/copying between files for it to associate those lines with the parent commit. And the default value is 40. If there are more than one -C options given, the argument of the last -C will take effect.
I have never once thought I know git.
me neither. I just use the basics and pray nothing goes wrong.
So excited to share this with my boss Mike, he's our local Git wizard and I'm so thankful for him. He's always so excited to share the new things he's learned and I'm looking forward to paying it forward!
Thanks for the book. It saved my neck in 2011. Every intern I've had since 2012, on their first day, I give them the book and ask them to read it all, and do all the exercises. I give them paper so they can mark it up as needed. Every intern that does this, has no issues with git all summer. The one or two interns that did not do the exercises? They struggle. Great book!
I haven't watched the video yet but it would be useful if you'd put book's name into your comment so people can look it up right away.
@@ghos7bear It is the "Pro Git" book you can find for free with a search engin of your choice
Pro Git 2nd ed. Edition by Scott Chacon -- just do every exercise and try every command, even if it's boring or you have done it before. That's what we learned over 10y of interns.
lmfao, you ask every intern to read a 500 page book for a summer internship? try not to respond with: "yeah I work at a super awesome place and im so smart, try putting a little more work in kiddo, pull up your bootstraps etc etc"
@@ronminnichhope you at least pay them for that lmao seems like a shitty workplace ngl. I have never had any problems with git and I didn’t get forced to read a 500 page book. If I wonder something I’ll just google it. Takes 3s
Didn't know that guy before, immediately caught fire because of his excitement, bought his book after the first 10 minutes. Great talk!
Time 8:13 - Setting or defining work and oss (open source software) directories. Thank you, Scott!
Really excellent talk. Incredible pacing, funny, and lots to learn for folks at every level.
Scott Chacon is a legend! I learned git in 2016, and I love it, it's one of the most important tools for me, after start my journey as a developer, I start to teach git, and the best resource always was the Pro Git book.
committing to multiple branches is brilliant! its always a PITA to notice/fix a core bug when working on a feature branch.. stash->checkout->pop->commit->checkout->merge is so much overhead for what started as a 5-second one liner fix.
This talk is great it felt I knew nothing of Github. But it was super fast delivery before my mind , eyes, and ears can align together and comprehend anything out of it
Fantastic speech! Thank you, Scott. So many cool things about Git.
"fucking nerds...". Golden
😂 family looked at me weird laughing my butt off
Indeed xd
Honestly, I find using it via the terminal much simpler lol
hahaha what a joke hahaha😲
My favorite part 😂✨
10:59 that -L to blame and log is awesome!
Best use for GUI Git has to be staging and unstaging parts of a file. In VS Code, just select some lines in the diff view and right-click. Stage selection and unstage selection. Amazing! 😊
you can do this in the terminal with `git add -p` - albeit not as elegantly when you want to break up a block of changes
Wow. What a great presentation. Thank you very much.
just pluggin git butler shortly, well done! Real-world git goodies, love it.
Virtual branches are an amazing feature.
There are so many times I've been working on a branch and seen a minor fix in a file unrelated to my work.
Streamlining those fixes while keeping them separate from my branches is very efficient for everyone.
Stash?!
For beginners who feel like god after knowing some git push command, this is a great humbling video for you.
Thanks for a great talk. Gonna be using git column all the time XD
46:30 I've been using Git for a long time and I use git gui for building my commit because I always commit lines of code, not full files. And I also use gitk for reviewing history of the code unless I'm looking for some highly specific thing that gitk doesn't support well - for those cases it's mostly `git log -p - - optional/path/limit` and grepping the results if needed. (The dashes should obviously be combined without a space but UA-cam comment interface cannot display that string correctly.)
I was there, great talk! Made me discover GirButler, super promising!!
I'm impressed you made the cut. I saw them ushering a hundred people out :)
Those packed aisles would have been a fire warden's nightmare.
@@jasonhenson7948 They kicked everyone out who did not have a seat before the talk started. We were sad, but the fire warden would have remained happy.
I was kinda hoping he would talk about git worktrees, I use those to checkout branches into separate file paths on my computer which allows me to work in different branches simultaneously. And I can keep long-lived branches separate. This also integrates really well with vscode workspaces.
frankly i have an old habit of sidestepping Git completely for that, i simply checkout the repo into two independent copies and i switch between those : )
Excellent talk! After many years on UA-cam this made me create a new playlist of videos I need to rewatch regularly. The bar has been set pretty high!
I wish you could search up, not based on the title, but on the frequency of words in the transcript of a video.
I loved your "Ask your code" talk back at Code.Talks and I love this one even more!
Holy shit that git butler idea is great, I've always wanted to work on several branches intermittently
Appreciated the Sneakers reference.
I was supposed to go to Fosdem, but family business took priority. Thanks for uploading!
Fantastic presentation!
Found this fun and informative... love the Sneakers reference too.
Waht a great talk. I have same frustrating about that part of git does not record all the shit for me.
I didn't know about range diff. I do the same thing manually because I didn't realise there was a tool for that :)
I realy love the pro git book and Im so happy that I found this vid 😊
Remember when version control used to be simple?
When it didn't need to be wrapped in huge frameworks so that people could work with it?
When it didn't need conferences like this?
When it wasn't the main topic of conversation on a project, and we just did the project?
Yes I do. And source control back then wasn't a tool the way git is. Like, also,git is awful in many ways, but I still love it deeply for the things it lets me do, manipulate and interrogate my repo history in many useful ways. It lets me approach certain problem classes in a meta way that old source control just could not. Branching is a first class citizen in git and that was not the case inb4. Learning git is a brick wall, as an early adopter I had the difficulty of explaining git many times to many companies, simple things are often stupid, and git basically requires you to learn it's mental model pretty well before you can do basic things with confidence. Then again, idk, sql doesn't make much sense unless you know what a relational database is. So git is hard in dumb ways, and if anything ever replaces it the way git ate svn, then I can't wait to drop git over whatever that future better system is, but it doesn't seem to exist, so I'm on the whole far far happier with git, warts and all, then any prior system. CVS wasn't even atomic, and vss would corrupt itself, svn really couldn't do branches and merge conflicts ruined your working copy in unrecoverable ways, and git is so so fast, and also best case was you could do exactly one local commit effectively where git you can do infinity local commits and also peer to peer stuff. Evolution of source control systems has been a net positive
@@rebeccakeller4666 Hi Rebecca. I could write a similar passage about how bad it is, but I won't bother. I will just say one thing; it is too difficult, and any good software should be easy. Yes, it can be learned, and when you have learned it, it is ok. But it shouldn't take so long to learn. I have been using Git for the last 10 years, so I feel like know it fairly in depth, but it took me about two years to get to this level. Why? You can of course take the easy way out at this point and just say I must be stupid, but I will assume you won't do that. It took me just a few weeks to get to a similar level with CVS, SVN, Clearcase etc. Now I know Git, but I have to work on projects with more junior people who are currently going through the same hell I went through. While they are learning, they make a lot of mistakes and people have to spend their time trying to fix them. The old systems, which would have improved a lot by now, were much better because they were simple and just worked. Anyway, I am glad you like it. Each to their own.
Yes, I remember SVN. It was simpler, but it came at a cost. For example working with branches was very clunky and inconvenient.
@@vibovitold Creating branches was just as easy as it is with Git. But you didn't need to make a new branch for everything. If it was a simple fix, you just checked out the code, did the fix, and checked it back in again, via a review if needed. I know I'm flogging a dead horse here, but Git is a huge step backwards.
@@jimmorrison2657 creating, kind of. merging, now that's not how i remember it at all.: )
and i'm not a fanboy of Git. it is poorly designed especially in terms of API - commands are poorly named etc.
but this actually emphasizes the question - if it beated all the competition DESPITE those blatant flaws... why did it happen in the first place? what's your theory
Youre a great speaker!
Thanks!
@@gitbutlerappYou did had a typo at least once ;P It's `git maintenance start` not maintainance. But I forgive you.
"Raise your hand if you never used anything other than Git" ... BRB feeling old.
Tell me about it.
@@gitbutlerapp 2x three letters: SVN, and FTP. 🥳
I feel actual anxiety when I recall the days of CVS and SVN. Man, those were horrible.
I still use RCS for personal projects involving a single file. For multi-file projects that I'm not making available publicly I'm more likely to use SVN . I started using git only because I was working on some Open Source Projects that used it. I don't like git that much. It is the only VCS that has resulted in me losing some work I had done but not commmited.
Svn and mercurial, thanks to mozilla
Great talk, I really enjoyed. I was wondering if the virtual branch is something different to what you can achieve with `git worktree add ` ? Amazing, helpful talk. Thanks
It is for the same problem, but an important difference. Worktrees give you a different working directory per branch, so if you have a feature and a bugfix, you don't have anywhere where you have _both_ the feature and the bugfix in one direcotry at the same time. With GitButler you do them both in the same directory and make branches from the hunks as though you were `git add -i` each hunk and committing.
In other words, we keep the work of the two branches separated in memory rather than on disk.
This is very nice if you want to see how branches will work together before they're technically merged in git.
@@gitbutlerapp thank you for the explanation. It makes sense and it's an awesome feature to have.
Wow, what a great talk! I'll add it to my collection of Git learning materials!
Git learning material???? REALLY? Aren't they more important things to learn in the whole world of IT????? LMAO
@@MJ-xg2ow there are many important things. But being dumb and unable to use major everyday tools is not the solution.
"Unable" to use major tools??? Wasting time to use git blame in cmd line to look cool while it's a click away is being Unable to learn everyday tool? I can't live without cmd myself but this is not the type of stuff one use cmd line to make their life easier....
@@MJ-xg2ow if it's available in your editor, it's fine. Some features are indeed more useful when visualized. But this doesn't mean that knowing fundamentals is useless.
Whoah that virtual branch stuff is exactly what I was looking for I expected you needed something like Pijul for that
This is the first time I heard of porcelain commands. A quick google and... it makes sense! Part of most software is plumbing that has to work, but is ugly, and not everyday user-facing code. The nice user-interfacing stuff is porcelain.
looking forward to gitbutler.
awesome talk!!
Hi, thanks for the session. FYI, you have a typo at 22:28
Namely `git maintainance` should be `git maintenance`.
"my voice is my passport, verify me" -- I love the reference! :D
Awesome talk! But I was looking forward to hear about "git worktree" too.
Very nice talk, I like his energy on the topic we all use all day!
Is there any source for the intro song?
43:10 - Working on multiple branches at the same time / multiple indexes, heads: Isn't that what `git worktree` does? I use it quite a bit, especially if I need to make a change to one branch while I have a bunch of other uncommitted changes in my main working tree.
I also use submodules - I have a lot of 3D printing projects using OpenSCAD, and I've developed a common library of functions / transformations / etc. Since OpenSCAD doesn't have a package management / dependency system like Python or NPM, I just add my common library as a submodule to each individual project.
Great talk! So many new gems to add to my workflow
I do a lot of code review and have been saved by reflog more than I care to admit 😅
I jump between branches a lot and sometimes mix up where I should commit or reset lol.
My friend's company is switching from SVN to Git after decades and the one piece of advice I told them was the reflog will save you from almost anything!
@@HtotheG I use stash a lot as well, can't be bothered to commit wip stuff all the time lol
I enjoy this talk. Thank you.
the GUI vs CLI debate, I never think about that anymore since I started using Magit.
You are hilarious!!!!!! In the positive way !!!!
🎯 Key Takeaways for quick navigation:
🛠️ Git configuration: Learn to set up aliases using `alias` and `git config` for better workflow customization.
🕵️ Git Blame: Utilize `git blame` with options like `-L` and `-w` to get detailed information about changes in specific lines and ignore whitespace.
🔄 Git Log Options: Explore options like `-S` for pickaxe search and `--word-diff` for word-based diffs in `git log`.
📜 Git Reflog: Leverage `git reflog` to view a log of reference updates, helpful for tracking changes, especially after operations like reset.
🔄 Reuse Recorded Resolution (`rerere`): Enable and use `rerere` to automatically handle recurring merge conflicts, saving time during rebases and cherry-picking.
🌐 New Git Branch Options: Discover the `--column` and `branch.sort` options for `git branch` to enhance the display of branch information.
🌟 Git Column Command: Learn about the new `git column` command, a simple tool to format input into columns.
🚫 Force with Lease: Familiarize yourself with the `force-with-lease` option, a safer alternative to force pushes, preventing accidental overwrites of others' changes.
🔄 Force with Lease: A safer option for force pushing in Git. It checks for stale references before pushing, making it useful for scenarios like rebasing or amending commits.
🔐 Signing Commits with SSH: Git now supports signing commits using SSH instead of GPG. This is beneficial for those facing issues with GPG and provides a similar level of security.
🛠️ Git Maintenance: Running `Git maintenance start` in your repository adds a cron job for background maintenance tasks, improving overall performance by optimizing tasks like commit graph generation and loose object collection.
📂 Large Repository Handling: Tips for handling enormous repositories, with insights from Microsoft's efforts in making Git work well for large projects like Windows. This includes features like prefetching, commit graph, file system monitoring, and partial cloning.
🏗️ Sparse Checkout: Useful for monorepos, sparse checkout allows you to clone specific directories, reducing the amount of data fetched and improving performance.
🔄 GitHub Merge Options: GitHub offers various merge types like merge commits, rebasing, and squash merging. You can enforce these standards and auto-merge pull requests based on CI status.
📝 Linear History Requirement on GitHub: GitHub allows you to enforce a linear history, rejecting pushes with merge commits. This is useful for maintaining a clean and straightforward project history.
🔄 Pull Requests with Heads: Pull requests on GitHub have corresponding refs, enabling you to pull down and merge changes directly, providing an alternative to using the GitHub interface for merging.
🔄 GitHub allows fetching from various refs directly, not just adding remotes for pull requests.
🌿 GitButler introduces virtual branches, enabling simultaneous work on multiple branches without switching.
🚀 Virtual branches in GitButler facilitate handling features and bug fixes separately without commits into the feature branch.
❓ A question is raised about Git range diff, with the speaker suggesting it's not widely used in typical workflows.
🔄 Submodules are criticized for their limitations, and the trend is shifting towards mono repos for better project management.
🛠️ Suggestions for Git improvement include the ability to work on multiple branches simultaneously and automatic recording of all changes.
🚀 The speaker discusses their preferences for using both command line and GUIs, depending on the task, with command line being faster for most operations.
🤔 Future directions for Git are explored, with emphasis on improving the user experience for modern workflows, such as multi-branch development and comprehensive change tracking.
Really nice talk. Thanks :)
really an amazing speech. thanks
Working on multiple branches can be achieved to a certain extent by using worktree command.
It will give you the ability to have multiple worktrees pointing to different branches، and the ability to switch between them while keeping the changes.
It isn't the easiest to use, but better then nothing.
only thing I know for sure that was added recently was more a change, the default from 'master' to 'main' (about time!)
22:17 His excitement is priceless.
I love the interactive add mention - seriously one of the three things I use `tig` for.
Because a piece of software is bloated doesn't mean I need to bloat my mind with it's bloat
I don't think I know Git. I just have enough knowledge to use it at a level that fulfills my needs
Awesome intro 👏
Great lecture. Could you please share what command you used at 7:50 as bb alias to print this nice version of the branches
This is too nerd even for me, but I’m really glad they all exist.
Fantastic talk!
The recording feature would be great. We've all been there, when we did a lot of work and then accidentally deleted the code, and it wasn't added/committed yet. And you pull your hair out wonder why th you didn't commit it... and you have to rewrite everything. It would be great if git allowed you to undelete something.
4:02 ahhh man, the plumbing commands was what I was REALLY hoping to hear more about..
I often like to temp-merge a big selection of branches together, write that to a tree, then commit the tree with a massive bunch of parents, of where I merged from.. Give myself a giant octopus merge and then throw away all the temp-merging commits, so you have a nice clean "MERGE ALL THE FEATURES AT ONCE" commit.. Looks fantastic :D
Merge does do octupus with porcelain, not sure what you mean
Hehe. Starting to watch the video. I am using git for 15 years (and before that hg, perforce, svn, cvs), and I would say I do not know git even on the surface. I know more about git in my company than anybody else probably, but I still would say I do not know git. And I do not want to know it.
I know we're doing a git client, but I do have to say that git is amazing :)
Same here plus sourcesafe, clearcase and tfs. Used hg for many years, wrote many extensions for it, even up-streamed some fixes in the early days when windows support was shaky. Git is the best by far.
13:01 WOW! That's the way to use blame!
Great session
What a great talk. Thanks :)
The fact that a nearly hour long video like this about Git even exists really shows why Git needs to be replaced by something better, something intuitive.
Start by not naming commands that do different things with synonyms lol.
Git is the JS of source control systems.
The naming is horrible, but the list of Git wrappers trying to provide a more friendly facade over its API longer than my arm, and yet none of them have ever gained much traction.
Of course it doesn't mean they aren't legitimate improvements.
I would liken it to the QWERTY layout. It's demonstrably not an optimal layout, quite poor in fact, but inertia is a tremendous force, so despite -perhaps precisely because - the number of alternative layouts is huge, none of them has really caught on.
@@vibovitold Yeah but you have to factor in the pain level and/or benefit in alternatives.
Is there really a significant benefit to a different keyboard layout? Most people don’t even touch type much less do it all day long without lots of pauses. So gains for most is zero to close to zero.
I think the Git tools trying to make it intuitive are pretty common. Every IDE of note has at least one, then there’s a lot of stand alone ones.
Most devs aren’t using Git on the command line for sure. But even with these tools you’re still dealing with an over complicated change control system that just wasn’t designed for internal teams but for loosely collaborating groups working on OSS.
For doing day job work at companies and gov orgs, it’s just not the right tool for the job but we got stuck with it till some new trendy tool comes out.
Trying to learn Git has destroyed my brain. It's shocking.
@@vibovitoldI use TorsoiseGit. It makes the impossible possible. I would have gone insane without it.
@@vibovitold That's not a fair comparison. There are very few people who are dissatisfied enough with QWERTY to push for a replacement. Most people seem so neutral on the topic it'll probably never cross their minds that an alternative could exist. The dissatisfaction with Git is much more tangible: I've never even used another version control tool and I still feel something better than Git should exist.
Great video
Never thought, that there were tools properly managing deconfliction, while everyone is telling about His obvious insight.
thank youuuuuu I just instantly subscribed
Thanks for the talk, many new things. I will try to incorporate some of them in my projects.
P.S. He looks like pedro pascal : )
Most important command I learned *git blame*
Nice talk, thanks
thanks for saving me uncountable hours!
Great talk, are the commands listed somewhere for copy/pasting?
What a fun talk!
I git so much good info there.
Great video thanks 🎉😊
Wicked fkin opening soundtrack, bombaclaat!
Neat. Learned a few things. Thanks.
Excellent!
Our lead dev still makes everyone manually copy-paste code between branches because "it's better this way, trust me".
At 43:04 you said you'd like to be able to work on more than one branch at a time. What are your thoughts on git worktrees?
they're fine, but they're not together. "at the same time" meaning with both branches applied to the same directory. like being merged but still being able to work on them independently. this is what we do with GitButler with virtual branches.
Awesome talk!
Had I beenin the audience, I would have asked this: I am a rebase guy, I organically hate merging master/main into my feature branch. This strategy works great as long as I work on a solo project. The moment a conversation starts and comments/requests to your changes land in, the problem kicks in. When you rebase to refresh your master/main, then git push force, the comments usually get misplaced, the refs shift, and, most of all, your reviewers lose track of what has been changed since their last visit. Is there a golden solution to that?
Nope. Unfortunately the best way to preserve history is by not rewriting it. You could squash merge so the whole branch becomes a single merge and try to incorporate comments as commit data, but otherwise it's quite difficult unless you just merge.
@@gitbutlerapp Thanks for your reply. I'm at least glad I'm not overlooking anything and it's just the way it is. I hoped though, there was a piece of AI that was able to bind the overwritten history with comments ;-)
Rebasing is better for going main->feature, and merge is better for going feature->main. The more authors are contributing to a branch, the less you want to rebase it, and the better an explicit merge commit is for tracking. IMO.
I do rebase -i for clean history all the time. Only rebase unpushed stuff. Problem solved 🤓
@@LA-MJ Here is the problem. If someone during PR asks you for changes then you cannot really git rebase -i anymore and group the changes with the commits where they belong. So no, the problem is not solved, your history is only neat until the original commits have been pushed, all that follows is still a mess.
I solved the millennium problems and quantum mechanics before learning git
"my voice is my passport" - haha, sneaky Sneakers reference... loved that movie 😀
I assume why the filter for clone takes "blob:none" for no blobs but "tree:0" for trees is because the first means "no blobs", but the other doesn't mean "no trees" but "root of tree", i.e. index 0.
Squash and merge ✊🏾
_This is the way_
FOSDEM appears a really cool meet event :) It's a while since I saw so many "F@cking nerds" in one place 😁
There are many more in other talks that happen simultaneously. That is the fun/dilemma of Fosdem
If anyone has any questions or suggestions for the next version of this talk, we would love to hear them!
More about CRDT stuff :)
Close the cameraman in a closet before next talk.
On a serious note, stacked PRs could be a nice topic.
@@ivanbessarabov690 interesting, we could do more on that. :)
I’m at half and you haven’t made a demo of *git-bisect* with a 300GB repository yet ;)