This is not the first time I’ve seen memes like this, and it makes me so glad I’m not involved in programming or software development. I would straight up die.
If you can find a comfy mid level role or a “real” senior role that is mostly code it’s a very rewarding career. But yeah I’d lose it with day long meetings
In one notorious incident, Warren added a DWIM feature to the command interpreter used at Xerox PARC. One day another hacker there typed delete *$ to free up some disk space. (The editor there named backup files by appending $ to the original file name, so he was trying to delete any backup files left over from old editing sessions.) It happened that there weren’t any editor backup files, so DWIM helpfully reported *$ not found, assuming you meant ‘delete *’. It then started to delete all the files on the disk!
No doubt. git rebase is like a very sharp knife. In the right hands, it can accomplish great things, but in the wrong hands, it can also spell disaster.
As someone who HAS used it a fair amount, I generally don’t even recommend it to people unless they’re already VERY comfortable with the rest of git and ideally have some sense of how it works internally.
Yeah it is something people should take time to learn. I do think its “dangers” are pretty overstated, though, especially if you always do git rebase --interactive, since if anything goes wrong, you can easily get out with git rebase --abort.
In general there’s a pretty weird fear that you can fuck up git to the point at which you can’t recover. Basically the only time that’s really actually true is if you somehow lose uncommitted work in your working tree. But if you’ve actually committed everything (and you should always commit everything before trying any destructive operations), you can pretty much always get back to where you were. Commits are never actually lost.
You can get in some pretty serious messes, though. Any workflow that involves force-pushing or rebasing has the potential for data loss… Either in a literally destructive way, or in a “Seriously my keys must be somewhere but I have no idea where” kind of way.
When most people talk about rebase (for example) being reversible, what they’re usually saying is “you can always reverse the operation in the reflog.” Well yes, but the reflog is local, so if Alice messes something up with her rebase-force-push and realizes she destroyed some of Bob’s changes, Alice can’t recover Bob’s changes from her machine-- She needs to collaborate with Bob to recover them.
Pretty much everything that can act as a git remote (GitHub, gitlab, etc.) records the activity on a branch and makes it easy to see what the commit sha was before a force push.
But it’s a pretty moot point since no one that argues in favor of rebasing is suggesting you use it on shared branches. That’s not what it’s for. It’s for your own feature branches as you work, in which case there is indeed very little risk of any kind of loss.
Ah, you’ve never worked somewhere where people regularly rebase and force-push to master. Lucky :)
I have no issue with rebasing on a local branch that no other repository knows about yet. I think that’s great. As soon as the code leaves local though, things proceed at least to “exercise caution.” If the branch is actively shared (like master, or a release branch if that’s a thing, or a branch where people are collaborating), IMO rebasing is more of a footgun than it’s worth.
You can mitigate that with good processes and well-informed engineers, but that’s kinda true of all sorts of dubious ideas.
Pushing to master in general is disabled by policy on the forge itself at every place I’ve worked. That’s pretty standard practice. There’s no good reason to leave the ability to push to master on.
There’s no reason to avoid force pushing a rebased version of your local feature branch to the remote version of your feature branch, since no one else should be touching that branch. I literally do this at least once a day, sometimes more. It’s a good practice that empowers you to craft a high-quality set of commits before merging into master. Doing this avoids the countless garbage fix typo commits (and spurious merge commits) that you’d have otherwise, making both reviews easier and giving you a higher-quality, more useful history after merge.
Why should no one be touching it? You’re basically forcing manually communicated sync/check points on a system that was designed to ameliorate those bottlenecks
No, you divide work so that the majority of it can be done in isolation and in parallel. Testing components together, if necessary, is done on integration branches as needed (which you don’t rebase, of course). Branches and MRs should be small and short-lived with merges into master happening frequently. Collaboration largely occurs through developers frequently branching off a shared main branch that gets continuously updated.
Trunk-based development is the industry-standard practice at this point, and for good reason. It’s friendlier for CI/CD and devops, allows changes to be tested in isolation before merging, and so on.
The other half are about useful things, like what to do next, how your interfaces will look like, and “if you need help, just tell me, I can escalate it”.
Merge gives an accurate view of the history but tends to be “cluttered” with multiple lines and merge commits. Rebase cleans that up and gives you a simple A->B->C view.
Personally I prefer merge because when I’m tracking down a bug and narrow it down to a specific commit, I get to see what change was made in what context. With rebase commits that change is in there, but it’s out of context and cluttered up with zillions of other changes from the inherent merges and squashes that are included in that commit, making it harder to see what was changed and why. The same cluttered history is still in there but it’s included in the commits instead of existing separately outside the commits.
I honestly can’t see the point of a rebased A->B->C history because (a) it’s inaccurate and (b) it makes debugging harder. Maybe I’m missing some major benefit? I’m willing to learn.
I feel the opposite, but for similar logic? Merge is the one that is cluttered up with other merges.
With rebase you get A->B->C for the main branch, and D->E->F for the patch branch, and when submitting to main you get a nice A->B->C->D->E->F and you can find your faulty commit in the D->E->F section.
For merge you end up with this nonsense of mixed commits and merge commits like A->D->B->B’->E->F->C->C’ where the ones with the apostrophe are merge commits. And worse, in a git lot there is no clear “D E F” so you don’t actually know if A, D or B came from the feature branch, you just know a branch was merged at commit B’. You’d have to try to demangle it by looking at authors and dates.
The final code ought to look the same, but now if you’re debugging you can’t separate the feature patch from the main path code to see which part was at fault. I always rebase because it’s equivalent to checking out the latest changes and re-branching so I’m never behind and the patch is always a unique set of commits.
For merge you end up with this nonsense of mixed commits and merge commits like A->D->B->B’->E->F->C->C’ where the ones with the apostrophe are merge commits.
Your notation does not make sense. You’re representing a multi-dimensional thing in one dimension. Of course it’s a mess if you do that.
Your example is also missing a crucial fact required when reasoning about merges: The merge base.
Typically a branch is “branched off” from some commit M. D’s and A’s parent would be M (though there could be any amount of commits between A and M). Since A is “on the main branch”, you can conclude that D is part of a “patch branch”. It’s quite clear if you don’t omit this fact.
I also don’t understand why your example would have multiple merges.
Here’s my example of a main branch with a patch branch; in 2D because merges can’t properly be represented in one dimension:
<span style="color:#323232;">M - A - B - C - C'
</span><span style="color:#323232;"> /
</span><span style="color:#323232;"> D - E - F
</span>
The final code ought to look the same, but now if you’re debugging you can’t separate the feature patch from the main path code to see which part was at fault.
If you use a feature branch workflow and your main branch is merged into, you typically want to use first-parent bisects. They’re much faster too.
You’re right, I’m not representing the merge correctly. I was thinking of having multiple merges because for a long running patch branch you might merge main into the patch branch several times before merging the patch branch into main.
I’m so used to rebasing I forgot there’s tools that correctly show all the branching and merges and things.
The thing is, you can get your cake and eat it too. Rebase your feature branches while in development and then merge them to the main branch when they’re done.
I would advocate for using each tool, where it makes sense, to achieve a more intelligible graph. This is what I’ve been moving towards on my personal projects (am solo). I imagine with any moderately complex group project it becomes very difficult to keep things neat.
In order of expected usage frequency:
Rebase: everything that’s not 2 or 3. keep main and feature lines clean.
Merge: ideally, merge should only be used to bring feature branches into main at stable sequence points.
Squash: only use squash to remove history that truly is useless. (creating a bug on a feature branch and then solving it two commits later prior to merge).
History should be viewable from log --all --decorate --oneline --graph; not buried in squash commits.
Folks should make sure the final series of commits in pull requests have atomic changes and that each individual commit works and builds successfully alone. Things like fixup commits with auto squash rebase. THIS WAY you can still narrow it down to one commit regardless of the approach.
Merge keeps the original timeline. Your commits go in along with anything else that happened relative to the branch you based your work off (probably main). This generates a merge commit.
Rebase will replay all the commits that happened while you were doing your work before your commits happen, and then put yours at the HEAD, so that they are the most recent commits. You have to mitigate any conflicts that impact the same files as these commits are replayed, if any conflicts arise. These are resolved the same way any merge conflict is. There is no frivolous merge commit in this scenario.
TlDR; End result, everything that happened to the branch minus your work, happens. Then your stuff happens after. Much tidy and clean.
Yes. My rule of thumb is that generally rebasing is the better approach, in part because if your commit history is relatively clean then it is easier to merge in changes one commit at a time than all at once. However, sometimes so much has changed that replaying your commits puts you in the position of having to solve so many problems that it is more trouble than it is worth, in which case you should feel no qualms about aborting the rebase (git rebase --abort) and using a merge instead.
I have the bad habit of leaving checkpoints everywhere because of merge squash that I am trying to fix. I think that forcing myself to rebase would help get rid of that habit. And the good thing is that I am the sole FW dev at work, so I can do whatever I want with the repos.
Yes. They do. A lot of people will use vacuous terms like “clean history” when arguing for one over the other. In my opinion, most repositories have larger problems than rebase versus merge. Like commit messages.
Also, remember, even if your team/repository prefers merges over rebases for getting changes into the main branch, that doesn’t mean you shouldn’t be using rebase locally for various things.
How would rebasing my own branch work? Do I rebase the main into my branch, or make a copy of the main branch and then rebase? I have trouble grasping how that would work.
Say I work on authentication under feature/auth Monday and get some done. Tuesday an urgent feature request for some logging work comes in and I complete it on feature/logging and merge clean to main. To make sure all my code from Monday will work, I will then switch to feature/auth and then git pull --rebase origin main. Now my auth commits start after the merge commit from the logging pr.
You’re still rebasing your branch onto main (or whatever you originally branched it off of), but you aren’t then doing a fast forward merge of main to your branch.
The terminology gets weird. When people say “merge versus rebase” they really mean it in the context of brining changes into main. You (or the remote repository) cannot do this without a merge. People usually mean “merge commit versus rebase with fast forward merge”
You nailed it with the critique of commit messages. We use gitmoji to convey at-a-glance topic for commits and otherwise adhere to Tim Pope’s school of getting to the point
Quasi parallel reply to your other post, this would kind of echo the want for a capital letter at the start of the commit message. Icon indicates overall topic nature of commits.
Lets say I am adding a database migration and my commit is the migration file and the schema. My commit message might be:
<span style="color:#323232;"> 🗃️ Add notes to Users table
</span>
So anyone looking at the eventual pr will see the icon and know that this bunch of work will affect db without all that tedious “reading the code” part of the review, or for team members who didn’t participate in reviews.
I was initially hesitant to adopt it but I have very reasonable, younger team mates for whom emojis are part of the standard vocabulary. I gradually came to appreciate and value the ability to convey more context in my commits this way. I’m still guilty of the occasionally overusing:
<span style="color:#323232;"> ♻️ Fix the thing
</span>
type messages when I’m lazy; doesn’t fix that bad habit, but I’m generally much happier reading mine or someone else’s PR commit summary with this extra bit of context added.
I see things like adding dependencies but I would add the dependency along with the code that’s using it so I have that context. Is the Gitmoji way to break your commits up so that it matches a single category?
Yes, that is another benefit, once you start getting muscle memory with the library. You start to parcel things by context a bit more. It’s upped my habit of discrete commit-by-hunks, which also serves as a nice self-review of the work.
I don’t see that as a benefit tbh - if I have a dependency, I want to see why it’s there as part of the commit. I’m imagining running blame on Cargo.toml and seeing “Add feature x” vs “Add dependency”. I guess the idea is it’s “➕ Add dep y for feature x” but I’d still rather be able to see the related code in the same commit instead of having to find the useful commit in the log.
I suppose you could squash them together later, but then why bother splitting it out in the first place?
I see that some use a subset of Gitmoji and that does make sense to me - after all, you wouldn’t use all of them in every project anyway, e.g. 🏷️ types is only relevant for a few languages.
I must have read that blog post in the past because that’s exactly the style I use. Much of it is standard though.
One MAJOR pet peeve of mine (and I admit it is just an opinion either way) is when people use lower case letters for the first line of the commit message. They typically argue that it is a sentence fragment so shouldn’t be capitalized. My counter is that the start of sentences, even fragmented ones, should be capitalized. Also, and more relevant, is that I view the first line of the commit more like the title of something than a sentence. So I use the Wikipedia style of capitalizing.
100% they do. Rebase is an everyday thing, merge is for PRs (for me anyway). Or merges are for regular branches if you roll that way. The only wrong answer is the one that causes you to lose commits and have to use reflog, cos…well, then you done messed up now son… (but even then hope lives on!)
Never use rebase for any branch that has left your machine (been pushed) and which another entity may have a local copy of (especially if that entity may have committed edits to it).
I never looked into this, so I have some questions.
Isn’t the overhead of a new function every time going to slow it down? Like I know that LLVM has special instructions for Haskell-functions to reduce overhead, but there is still more overhead than with a branch, right? And if you don’t use Haskell, the overhead is pretty extensive, pushing all registers on the stack, calling new function, push buffer-overflow protection and eventual return and pop everything again. Plus all the other stuff (kinda language dependent).
I don’t understand what advantage is here, except for stuff where recursive makes sense due to being more dynamic.
Some languages have to optimize it with various tricks. There’s a good reason why I call heavily functional “programmer wankery”. It took me a while to run into an issue that was caused by a variable modified in a wrong way, which I fixed by saving the value of the variable before a call that seems to alter it. Probably I should have instead properly fix it so I could understand the actual root cause, but I have limited time to spend on things.
They aren’t talking about using recursion instead of loops. They are talking about the map method for iterators. For each element yielded by the iterator, map applies a specified function/closure and collects the results in a new iterator (usually a list). This is a functional programming pattern that’s common in many languages including Python and Rust.
This pattern has no risk of stack overflow since each invocation of the function is completed before the next invocation. The construct does expand to some sort of loop during execution. The only possible overhead is a single function call within the loop (whereas you could have written it as the loop body). However, that won’t be a problem if the compiler can inline the function.
The fact that this is functional programming creates additional avenues to optimize the program. For example, a chain of maps (or other iterator adaptors) can be intelligently combined into a single loop. In practice, this pattern is as fast as hand written loops.
A great point in favour of maps is that each iteration is independent, so could theoretically be executed in parallel. This heavily depends on the language implementation, though.
Imperative for loops have no guarantee at all that iterations could be executed in parallel.
You can do some (usually expensive, and never complete) analysis to find some cases, but smart compilers tend to work the best the dumbest you need them to be. Having a loop that you can just blindly parallelize will some times lead to it being parallel in practice, while having a loop where a PhD knows how to decide if you can parallelize will lead to sequential programs in practice.
Compiler optimizations like function inlining are your friend.
Especially in functional languages, there are a lot of tricks a compiler can use to output more efficient code due to not needing to worry about possible side effects.
Also, in a lot of cases the performance difference does not matter.
I’m not familiar with any special LLVM instructions for Haskell. Regardless, LLVM is not actually a commonly used backend for Haskell (even though you can) since it’s not great for optimizing the kind of code that Haskell produces. Generally, Haskell is compiled down to native code directly.
Haskell has a completely different execution model to imperative languages. In Haskell, almost everything is heap allocated, though there may be some limited use of stack allocation as an optimization where it’s safe. GHC has a number of aggressive optimizations it can do (that is, optimizations that are safe in Haskell thanks to purity that are unsafe in other languages) to make this quite efficient in practice. In particular, GHC can aggressively inline a lot more code than compilers for imperative languages can, which very often can eliminate the indirection associated with function calls entirely. gitlab.haskell.org/ghc/ghc/-/…/generated-code goes into a lot more depth about the execution model if you’re interested.
As for languages other than Haskell without such an execution model (especially imperative languages), it’s true that there can be the overhead you describe, which is why the vast majority of them use iterators to achieve the effect, which avoids the overhead. Rust (which has mapping/filtering, etc. as a pervasive part of its ecosystem) does this, for example, even though it’s a systems programming language with a great deal of focus on performance.
As for the advantage, it’s really about expressiveness and clarity of code, in addition to eliminating the bugs so often resulting from mutation.
So it basically enables some more compiler magic. As an embedded guy I’ll stay away from it, since I like my code being translated a bit more directly, but maybe I’ll look into the generated code and see if I can apply some of the ideas for optimizations in the future.
I looked at the post again and they do talk about recursion for looping (my other reply talks about map over an iterator). Languages that use recursion for looping (like scheme) use an optimization trick called ‘Tail Call Optimization’ (TCO). The idea is that if the last operation in a function is a recursive call (call to itself), you can skip all the complexities of a regular function call - like pushing variables to the stack and creating a new stack frame. This way, recursion becomes as performant as iteration and avoids problems like stack overflow.
Not just calls to self - any time a function’s last operation is to call another function and return its result (a tail call), tail call elimination can convert it to a goto/jump.
programmer_humor
Active
This magazine is from a federated server and may be incomplete. Browse more on the original instance.