I started using Linux in 1997. I've been working at Microsoft for a little over 4 years. There are some annoying things about Powershell (verbosity and slow loading of tab completions etc), but in some ways it's vastly superior to bash, as it deals with structured data and not just strings you have to parse, and it has an extremely powerful standard library API (.NET) at your fingertips.
I still love bash and use it as well. But the hate powershell often gets is misguided.
There was a time in Powershell’s life when stdout redirection to a file (>) would write UTF-16, then that changed to UTF-8 with BOM and now it’s UTF-8 no BOM. During those times some PS versions would allow to change this globally, some would not. Even if you could change it globally, the names for the same encodings have changed.
PowerShell 5.1 “UTF8” actually meant “UTF-8 with BOM”
PowerShell 7 “UTF8” meant no BOM.
If you read a file in PowerShell that does not have a BOM it decodes it as ANSI. Yes, ANSI, not Ascii. Which is a hilarious choice because you can’t even write an ANSI file in PowerShell. The closest encoding you can use is “Default” which will use the system default but your systems default might not be ANSI.
PowerShell is a hot mess. Go read the official docs for character encoding[1] if you don’t believe me.
I have CI scripts that have to use .NET calls[2] in PowerShell to write ssh keys to disk because it’s impossible to write utf-8 no BOM (or ascii) with out carriage returns (\r) being inserted before every new line (\n) even when the incoming string didn’t have them.
File encoding is just ONE of the reasons I hate PowerShell and why it’s so obviously clear people who were not qualified to design a shell, designed a shell.
I'm not sure if we should blame the designers of PowerShell (Jeffrey Snover and his team). If you followed the history of their project, it seemed like they had very good ideas, but it was very hard to get buy-in for anything CLI-oriented in Longhorn/Vista-era Microsoft[1]. They've all had experience with Unix shell and their ideas for object based pipes were truly innovative as far as I know.
I can't speak for Jeffrey and his team, but I feel like a lot of their decisions came from trying to get corporate behind the shell and present it as a shell for Windows. They avoided picking political battles outside of their main goals (a modern shell to replace cmd.exe and the object-based pipeline model). What we've got are a set of decisions that aligned with Windows and Microsoft practices of that day and age:
Microsoft is focusing on .Net as their general platform? We'll implement PowerShell on .Net.
Windows's standard for Unicode text files is either UTF-16 LE or having a BOM for any other Unicode transformation? We'll do UTF-16 by default and always add a BOM if you choose UTF-8.
Windows is using CRLF? Well, we'd pipe CRLF-delimtied text by default.
Visual Basic and C# programmers expect functions to have names like "GetChildItem" instead of something like "dir" or "ls"? No problem, we will set the canonical command name to be a long, programming-language-like name and set up aliases that look (but don't behave) like Unix and cmd.exe commands.
The result was not pretty, but I still appreciate the ideas we got from PowerShell. nushell took these ideas and implemented them in a more modern way.
I work regularly in PowerShell and this bugged tf out of me. The result of my digging was: .NET libraries emit textual output in <whatever the offending encoding was, UTF-16 LE or something> -- and since PowerShell is implemented in .NET, it necessarily inherited that default.
Things have improved since then now that everything is implemented in .NET (Core), so if you're working with PoSH 7.x, this is no longer a headache. (I just tested in 5.1, and it seems fixed now, as well.)
Of course, by now, I've just developed muscle memory to add '-enc UTF8' to the end of everything that writes text to disk.
> File encoding is just ONE of the reasons I hate PowerShell and why it’s so obviously clear people who were not qualified to design a shell, designed a shell.
I find that unfair towards their creators. They invented a truly innovational shell, whereas everybody else continued the text-based approach. The object-oriented approach is so much simpler when it comes to process results of Get-ChildItem (ls) or Get-Process (ps) when you get objects with properties, instead of just text. Especially when in text-based shells it matters how you call e.g., ps (aux or -efH or whatever).
Now, I find your statement unfair, because in Linux world, you just create a new shell and whoever wants to use it, can.
At Microsoft, a successor to cmd.exe had to be shipped with Microsoft, otherwise it would never have been adopted. Most big companies would never allow a third-party open-source shell on their Windows servers. Therefore, you must navigate a big ocean of politics and powers, guarantee nackwards-compatibility and meet expectations of thousands of companies. This inevitably leads to behaviour like encoding that is frustrating to use. Until you read the docs, which state quite comprehensively what you have to expect - as you discovered yourself.
It's also worth pointing out that those Microsoft docs differentiate between the two major versions of powershell using the names "Windows Powershell" and "Powershell".
The original project name was "Monad", but once it was nearing release, it got into the hands of the notorious naming prodigies in Microsoft marketing. Then it got changed to Microsoft Shell (MSH) and finally to Windows PowerShell. Great times.
Windows Powershell is up until 5.1 because it runs ONLY on Windows.
Powershell 6 was named Powershell Core due to being built with .NET Core. It was not fully compatible with 5.1 and lacked some functions based on the WinApi.
PowerShell 7 is still built on .NET Core, but has a much better compatibility with 5.1. They dropped Core because nobody cared and it's shorter and works equally well on all platforms.
So Powershell is the v7 and Windows Powershell is the old 5.1. Nowhere is that confusing.
Also worth noting that PowerShell is completely independent from Windows now. Not only can you install it in almost every other platform, since it's .NET you can embed the runtime in any .NET application. It's incredibly powerful.
Not sure what you mean? I've hosted powershell in C# code by referencing Microsoft.Powershell.Sdk - it pulls in the dll at compile time and runs on whatever .Net runtime that is installed and matches the project target.
Wait - you might just have helped me to fix my Invoke-WebRequest testing 3rd party api we have to integrate with. Unexpected character encountered while parsing error line 0 position 0 seems like there is a BOM I wasn't expecting to send in my request and receiver definetly was not expecting to receive it.
Edit: yup, just did small cmd line client and all works. Thanks.
UTF8 with bom is a terrible idea. “Why does the name of the first column in my csv file not match?” is a question I answer at least once a month. It’s as if they’re trying to EEE plain text.
Hey let's not pretend like it's better in the world of Linux/Unix/etc. Now that is one hot mess of everything, just as much as Powershell. It's just those have things we're used to, and kinda learned to (mostly) abstract away without the presence of an over-bearing behemoth such as Microsoft which makes such a thing almost impossible.
As a comparison (to most of us semi-knowledgeable windows / powershell users), ask any semi-knowledgeable Linux user about the difference between Ksh/Bash (or that weird one that Ubuntu pushed and aliased to one of the above for extra confusion). Or ask them if any of them remember how tf to use SED or AWK without looking up the docs because those damn things are made by satan in the depths of hell. Or how to use parameters inside shell scripts, and the various incantations they have to slice and dice the params.
Nushell is plenty fast, works with structured data (also has a vi mode if you care, I do) and works equally well on Windows and Linux. You can install it with winget on windows and with cargo or the system's package manager on linux.
I am a C# developer so I should feel right at home with Powershell. In some ways I am. I do manage to get things done if I have to use it. If I am inside the Windows environment (with other Windows devs) I will try to encourage using it over CMD.
One thing I will give Powershell credit for is, back in 2012, being able to chuck something together when I did not have the development tools I needed. I knew that Powershell was .NET under the hood so I had a stab at being able to load my DLL files and calling methods inside a class. It did the trick! I was able to hand it over, suggesting they use it as a Task Scheduler (daily) and hardly anyone touched it during the remaining of that promotion (4 months)
Despite all this, if I ever have to write a script for odd execution or regular automation... I wish I could use bash. I have mentioned Git For Windows or using WSL but I just cannot win anyone over. Its not like anyone is an expert on shell scripting.. powershell or whatever.
As much as I like Powershell -- I still consider bash to be better. I think half of this reason boils down to what you are accustomed to. I have been using Linux properly since.. around.. 2006.
I am a C# developer and I know it's basically .Net as shell, but I have no clue what to do when in PowerShell. I'm always lost, and without searching the internet I get to nothing. In Bash/Zsh I quickly reach my goals.
I just think coming from a Linux background, it's very hard to wrap my mind around the mindset of PowerShell which is just quite different from Bash.
I've used powershell a bit, particularly for scripting CI pipeline steps. No matter how much I use it, I can never remember anything about the syntax. I primarily use C#, so you'd think there'd be some affinity, but nope. It's like a black hole.
I've written in a lot of different languages in my time and never had this issue with anything else. It's weird.
Yeah bash has some of the worst syntax of any language I've used. Why do if statements have a semicolon in the middle of them?? Why does it use `[[` and `]]`? why does `if` end with `fi ` but `while` ends with `done` ?
people who rip on powershell but prefer bash make no sense to me at all
Exact same here. I write roughly the same amount of Powershell as I do Bash, and yet to this day I feel like I barely know Powershell and constantly look up the simplest things, whereas loads of dark corners in Bash stick in my mind perfectly fine.
AIUI the verbosity is by design to make scripts more clear, and users are expected to use aliases to make cmdlets easier to work with interactively. But I assume that few of those aliases are actually standardized...
Powershell introducing the English language to a shell at a time when everyone else was doing i18n is about 3 leaps beyond an annoying thing in the traditional Microsoft form of the time.
You can do what ever you want for innovation but if you choose mindf*ck as your UI..
I may be misunderstanding, but you can write to the console host in a background task. To get an output stream, you need to await the task and collect the result.
I feel it's nice to have the power of choice, I can deal with structured data in bash as much as I need, whenever I need, with awk, but it isn't there when I don't
Bash rocks because it is about the user and the OS. Powershell is about one thing, Microsoft and the ugliness of ItS CumBerSomE-SynTaX Du-Jour is proof.
It's about what twenty people could sling together and make work on a Vax in the 80s in 4Kb.
Powershell has regular syntax, legibility, a high degree of discoverability, and included batteries for things that bash victims reimplement again and again.
It may not be the best shell for your purposes - it certainly won't run in 4Kb - but it wipes the floor with all legacy shells for usability.
You're welcome to have an aesthetic opinion on the syntax, but you must be dead inside if you can look at an average bash script and not recoil in horror. Argument parsing? sed? Control statements? Quoting and escaping?
"It's all about the user" is it fuck, you're on drugs.
> I’m more comfortable with bash commands than CMD or PowerShell commands (because I’ve been using Bourne shell (sh) and friends for more than 30 years).
This is the only argument I'm willing to accept, and honestly it's the only argument the writer needs, really: un-learning 30 years of experience to use something else even if the latter is objectively superior, would be a pain in the neck for anyone.
That being said I believe that seriously attempting to use the native tools would help the author understand Windows at a more fundamental level.
Also, this does raise a few questions, though—if the writer has been using UNIX shells for that long, why are they using Windows at all? As I commented elsewhere, why shoehorn the whole set of UNIX core utilities and binary utilities onto Windows when the latter comes with its own shell, command-line utilities, and management tools? More importantly, the author going by the rest of their posts (assuming it was the same author) doesn't strike me as the type (i.e. normie, for lack of a better word) to use Windows as a daily driver, but they seem to use it all the same. Therefore, why not an OS where these utilities feel more native and more useful, like Linux or one of the BSDs?
In contrast to the author I've been using and programming almost nothing but Windows (with a brief stint on Linux in university and now at work, but only for work), and on Windows I use PowerShell and if I really need to, CMD.exe (good riddance). As far as compilation goes, I use Visual Studio with MSVC and Clang-Cl, because they are the only complete development environment for Windows—all the others (MinGW, Cygwin, MSYS2) are incomplete or worse in some way, or make life harder than necessary in some way.
> I have experience using Windows PowerShell (powershell.exe), but do not yet have experience using the cross- platform PowerShell Core (pwsh).
For the record, functionality between powershell.exe and pwsh.exe is almost identical, with some minor differences[1]. Additionally they differ in the .NET runtime they were written against (.NET Framework 4.8 for powershell.exe, and .NET 7 and later for pwsh.exe).
> Also, this does raise a few questions, though—if the writer has been using UNIX shells for that long, why are they using Windows at all? As I commented elsewhere, why shoehorn the whole set of UNIX core utilities and binary utilities onto Windows when the latter comes with its own shell, command-line utilities, and management tools?
I'm a Linux user and developer who writes software targeting Windows.
Tools like Git Bash and MSYS2 let me write one set of build scripts that work across Windows, macOS and Linux, because all of the tools exist on the three platforms.
I don't have a need to understand Windows on a more fundamental level, nor do I want to. I just want to meet my users where they are at, which is Windows/macOS/Linux.
Fair point. I've edited my comment to show my impression that the author uses Windows as a daily driver, rather than merely 'for work', which I think makes the cause for my confusion a little clearer.
> Tools like Git Bash and MSYS2 let me write one set of build scripts that work across Windows, macOS and Linux, because all of the tools exist on the three platforms.
Powershell is cross platform too these days, which makes this kind of a moot point.
Powershell on Windows is great because there’s a module for everything. I’d rather parse structured data.
Powershell on Linux isn’t worth is because I end up having to parse everything as text anyway, completely negating any benefit I can come up with. It doesn’t feel as natural as the *nix tools to me.
For what it's worth, I've had the opposite experience, maybe because I work in a different domain. Personally, I still think using Powershell functions and treating lines as string objects is an improvement over bash and its mini-languages, anyway; and most standard Unix commands have Powershell close equivalents. Some Powershell expressions might be abstruse but so is, say, awk; but when you've learned awk you only know awk; whereas the Powershell technique you used (calculated fields in select-object, for example) is permanently useful. I do spend a lot of my time with various random APIs so Powershell is pretty joyful compared to (say) curl/jq or yq-ish stuff. There was a real adjustment period, of course, getting used to a new shell; but it was the best thing about my experiment with Windows (discovering a new shell that I now use on my Unix-like systems, because I did not end up wanting to adopt Windows).
So I'm curious what kinds of things you mean when you say you have to parse everything as text anyway. I suspect you've needed to a lot more of those kinds of tasks than I have.
Powershell's CSV, XML and JSON handing is my favourite Powershell feature. Completely agree with you that it beats a *nix shell and things like jq. If I was interacting with those all day I might be convinced to use it more.
What I mean by parsing as text was more about actual commands.
I wasn't expecting 1:1, but as an example, let's say I wanted to get a list of local users on a machine.
On Windows I can use "Get-LocalUser". It returns an object with properties you can filter, output with a format, etc. Basic stuff that saves me doing any thinking. I like that. It's amazing how column titles add to the user friendliness.
On Linux, Get-LocalUser doesn't exist, so I could run something like "getent passwd". It's only option is to return a screen of text. There's no structure yet, just lines and colons. Now I can (even though I shouldn't) grab fields with cut, or awk, and filter with grep, and maybe output something pretty at the end with column. What does the 5th column mean again?
So on the Linux side, I've never been very motivated to use it, since I often have to fall back to old ways anyway.
Thanks, that makes a lot of sense. I can see a lot of similar cases especially for sysadmin things (systemctl, ifconfig, etc.).
One thing that was nice about my adopting Powershell as my daily shell is that for the most part, I could just use the same shell utilities as usual:
getent passwd | cut -f6 -d':' | sort | uniq -c
Like, there's no problem running that in powershell. I know people have talked about "mixing" strings and objects but, I have to say that for me I've rarely run into a problem with it?
As you learn you can do a little more. Maybe you learn that group (group-object) can be used instead of the two commands at the end:
getent passwd | cut -f6 -d':' | group
And for me, the nice thing is that when you learn something like that, it's leverageable across all your cases. You learned something about your tool, not just about the passwd file.
Is this better than bash? I don't think it's worse:
getent passwd | %{ ($_ -split ':')[5] } | group
But then you might notice that your data is a CSV (C for colon in this case) and you might leverage some of the CSV-handling which works well in Powershell:
Now you're cooking with structured data again. I think familiarity makes this kind of thing come naturally. And since these commands are pretty discoverable, because they're more consistent, and tab-completion and command-line editing are so much better than bash. It's easy to see how the above can be made terser by stashing it into utility function or a hashtable in your profile (e.g. so you could do "getent passwd | fields passwd" or something like that.
Now, I'm not going to tell someone it's worth the growing pains and adjustments to switch. It's like switching keyboard layouts. And matters of taste might turn you off, understandable. But for me there have been real benefits, and on its merits, I do think it makes a better shell. And it would be nice (from my perspective) if the community did some more work around these use cases and making them nicer for people.
Oh, it's not dismissive at all. It's not just a real consideration, it's an overriding one. I switched my "daily driver" shell to powershell but I still write CI/CD scripts in bash, and docker entrypoints, and cloud-init userdata, and utilities, because it's niche enough to be too much to ask my coworkers and community folks to also switch.
I do wonder about the muscle memory thing. I think having to create scripts in bash keeps my hand in enough that I won't lose it too badly. At least I hope so. I compared switching shells to switching keyboard layouts; something I also did, and something where it's been some effort to retain enough muscle memory to not completely flail when presented with another computer.
Yeah, it's hard to do a direct comparison, because bash needs a bunch of other utilities to be useful and they also take up space. In theory you can pick and choose but woe be to you if you want to use bash without sed, awk, cut, etc.
But it's true that, like, you get help with Powershell but that might not matter on a system where you wouldn't choose to install man pages or something.
But they aren’t on windows, so they can’t be taken for granted there. You have to install git bash on windows.
What’s the difference between installing git bash on windows and installing powershell on Mac/Linux? I would argue that there is none.
And obviously it’s fine that you don’t want to use PowerShell, I don’t want to use it either. But personally I also dislike git bash on windows. It doesn’t feel like it belongs there, just like powershell doesn’t really belong on Linux/Mac.
If I had to write a build script for all three platforms I would just write them for their native environments. I wouldn’t be all that willing to add another tool to be installed just to save myself the effort of simple build script.
I can run an MSYS2 installer on a Windows VM once and be done with it. I will have to roll my own containers, for example, if I want to use Powershell elsewhere, versus just using vanilla or vendor-provided images.
For my use case, I don't see the need to bend my builds around Powershell just to use Powershell. This plays into my other point about not wanting to dig into Windows fundamentals when I don't need to.
> If I had to write a build script for all three platforms I would just write them for their native environments. I wouldn’t be all that willing to add another tool to be installed just to save myself the effort of simple build script.
This was my initial approach, but it became too much work to maintain separate build systems, especially when it came to making changes over time. One change becomes three, along with three new opportunities for things to break in three separate ways.
For what I'm doing, moving to Unix-y toolchains simplified builds and made it much easier to reason about it.
I totally get that your solution works for you and that is fantastic. I’m only questioning the pure logic behind it when you strip away opinion/preferences.
You picking one OS that you’re okay with modifying but the other one you’re not. If we inversed that preference and said that I’m willing to modify my Linux containers but unwilling to install software on my Windows VM, the logic is the same.
By the way, Microsoft publishes Linux images that already have PowerShell pre-installed.
So, technically, if your build scripts were 100% PowerShell you could have it where both Windows VMs and Linux docker images could be unmodified.
Since your build scripts sound overly complex, they could possibly benefit from a scripting language that’s got support for objects and object data pipelines, more advanced error handling, and more advanced object parsing than tools like sed and awk.
Just playing “devil’s advocate” here, the logic checks out.
>> What’s the difference between installing git bash on windows and installing powershell on Mac/Linux? I would argue that there is none.
Big, big difference. My Windows laptop is basically a terminal to the Linux development server. I can install anything (within reason) on Windows, can't install shit on the Linux server unless I beg the IT admin for days and eventually he just says "no". He did install the "mc" (Midnight Commander) utility on it though and made my life like 100x easier. If MC were bundled with git bash, that would be something I would cheer for.
And yeah, I use git bash too on Windows. For accessing the Linux shell through ssh when I remotely connect to the server and for convenience locally. Main workhorse are the "less" and "grep" utilities to examine logs. What am I goint to use otherwise, Notepad?
What will you use to view logs, Notepad? Well, you could use the #1 most popular text editor that Microsoft happens to develop, VSCode. Or a wide array of Windows GUI software that isn’t available on Linux, like Notepad++.
The fact that your IT department has those specific policies in place and is inflexible with your developers isn’t really relevant to this discussion. The truth is that if you wanted PowerShell on Linux it is trivial to install. Microsoft even publishes Linux images that already include PowerShell.
Powershell had potential, but it didn't pan out. It's annoying and very challenging to remember all the one-off flags.
With bash, you only need to learn 10 or 20 short commands and a few flag variants, over a gently long period of time. Then you're all set for 97% of cases likely to be encountered.
This is the opposite of how I see it. There's really nothing universal about flags, especially short ones, in different random utilities (what's -n mean? -f? file, filter or force?). Whereas in Powershell since there actually is some standardization, your knowledge about what -WhatIf or -Verbose do is actually pretty leverageable. And more powerful with the ability set default parameters and so forth. And the discoverability and help in Powershell is also standardized.
So I get why unfamiliarity would make it not worth it to switch from whatever shell you prefer, but I don't think I can see your point about flags.
I used bash in college and then didn’t use it for around 5 years. And then I used powershell.
And bash is still a lot better as a shell language. Powershell commands are long and annoying. If you remember all the aliases (most of which are annoyingly bash commands but hardly behave that way) then the parameters are long and annoying.
Stringing commands together requires too much API digging because objects mean you need to run a variety of incantations on the results to know what members you need to pull out and what data format they have, etc. in Bash it’s always strings and if you know a couple of useful string manipulation commands such as cut, that you probably know anyways for other reasons, you can get by with stringing most commands fairly trivially.
Powershell OTOH is a lot better when I’m trying to write an actual reusable script. But at that point why wouldn’t I just use something like Python anyways (or if it will be associated with a front end repo I will often just use nodejs since it’s definitely installed). Python has much more usable comparison, control flow, etc operators, much bigger libraries, better multi OS compatibility etc.
The only advantage really is using powershell commands/applications, so sometimes I will write my scripts in Powershell.
But in Linux I will also sometimes write my scripts in bash for the same reason, so it’s not really an advantage inherent to powershell, but the fact that it’s the “standard” of the OS.
I do like Powershell, but the problem with it is that it tried to be both a great scripting language and a great shell language, but the things that make it a good scripting language, such as verbose naming, objects, etc make it a worse shell language than shell languages designed to be shell languages first, and the things that it does to make it a useable shell language, makes it a worse scripting language than other scripting languages.
On average it may be the best language across all environments and use cases. The problem is that for any specific environment and use case it’s rarely one of the better languages.
I'm not sure I understand your point regarding having to look at the output of commands to know how to use them. At least with Powershell you have tab-completion suggesting your object members; you have to at least look at whatever flavor of semi-parsable text is coming from your command in bash to use cut or awk or whatever on it (and the semantics of what you're looking at are not discoverable, so you're likely to have to some API digging of your own). 'cut' and the like is sadly fragile for a number of reasons, and you generally won't discover them in advance (e.g. when the date "field" starts containing a year, sizes overflow or start being indicated with human-readable abbreviations after a threshold, which is exactly the sort of thing you can't tell by inspection). And bash's failure modes are really sharp.
To be honest, something like
$titled = gci *.md| ?{ (gc $_)[0] -like '# *' }
seems short and less error-prone than the bash equivalent. Not sure what it would be. Something like
titled=(); for file in *.md; do if head -1 "${file}" | grep -sq '^# .*'; then titled+=("${file}"); fi; done
I think there's a lot of little gotchas in there, and not a little "API digging" for options, though it's simple in concept.
> you have to at least look at whatever flavor of semi-parsable text is coming from your command in bash to use cut or awk or whatever on it
Having to parse things is definitely a pain, at least until you get good at it. But the critical thing is that all the output is right there on the screen, and often in a format that's at least somewhat designed to be parsed. In PowerShell you have to go diving through the object hierarchy. That would be OK I guess if things were intuitive and the help and documentation were great but that's often not the case. And often the API semantics are designed for a different language altogether (C#) and the things you have to do to consume the API in PowerShell are ugly.
Your "diving through object hierarchy" has a _consistent_ command called "Get-Member". Not the same where you need to dive through several hundred text output structures for different commands and/or read man page options. I am sorry, but your statement is factually false.
The simplest technique for analyzing the objects that a command returns is to pipe the output of that command to the Get-Member cmdlet. The Get-Member cmdlet shows you the formal name of the object type and a complete listing of its members.
This is a very succinct dissection of the problems but I will add one more: PowerShell's frustrating behavior where any return value of a function call is part of the enclosing function's return value if not explicitly swallowed. I understand what they were going for by analogy with other shells here, but in my opinion in a language like PowerShell (with actual return values) this is unintuitive and makes it really easy to introduce bugs. The whole object pipeline thing they were going for feels like an evolutionary dead end that we shouldn't be saddled with.
I agree, this is a place where the desired features are in tension. I will say that it took me a while to find it, and I don't find it limiting now that I understand it, but it was a real gotcha when I encountered it.
- I had 10+ years of experience in linux/unix development
- Career change to AAA game development which is predominantly Windows based
- Day to day do Unreal C++ developing in Windows with Windows toolchains
- Still use WSL (and thus bash) for everything else, including DevOps related work, cloud/container development work, and personal notetaking/productivity via emacs+org-mode.
So, some of us have reasons to co-exist in multiple OS's. I tend to write up scripts in Python if I can so that there's at least a chance that I can run them in both Windows + Unix environments.
I feel similarly - I first learned about the shell with Linux and all my career has been with Linux, but at home I use Windows. I am making an effort to be better at PowerShell but if I just want to do something quickly, I already know how to do it with bash and I just want to get on with my day.
Perhaps the company is otherwise all in on Microsoft's ecosystem. Perhaps they use Outlook, Word, Excel and similar apps where the Linux version doesn't exist and the Mac version is an afterthought. So IT forces everyone to have a Windows system, and then give only the developers a separate Linux machine.
At this point, this should be the industry standard. "Local" should be whatever you can SSH into. You can get as much hardware as you need, and you don't have to worry about anything private (i.e., customer information) and/or proprietary being on your lost / stolen laptop.
Not just poor wording but I think you misunderstood the authors context too. Some of his comments were comparing git bash to other POSIX layers like WSL, rather than Bash vs PowerShell.
> I think you misunderstood the authors context too
This I did not. The post was about Git Bash as a command-line shell for Windows and everything it entailed, I understood that and that's why I wrote the paragraph that followed.
I was limiting the scope of my comment to the author's familiarity on Bash versus CMD or PowerShell because I considered that the cornerstone argument of the entire post. Everything else that the author mentions is expected by default in a *sh-type shell on a Unix-like OS; there's nothing special there. People wouldn't blog about being able to use `grep`, `ls`, etc on Linux; it's just everyday business.
> > I think you misunderstood the authors context too
> This I did not. The post was about Git Bash as a command-line shell for Windows and everything it entailed, I understood that and that's why I wrote the paragraph that followed.
I literally just said he wasn’t just comparing Git Bash with Powershell though. So you can’t have understood their context if you took Powershell to be the “cornerstone”.
It’s a bit of a muddled blog because he is constantly switching focus between “Git Bash” (the package / distribution; whatever term you want to describe it as) and Bash.exe (the shell).
There are a lot of the comparisons are with things like WSL, for example when he discusses the underlying file system. And when he talks about the ease of installing and keeping things up to date. None of that was directed at Powershell specifically, nor even at all.
I said the cornerstone argument for their blog was 'they were familiar with Bash, so they don't use the shells that come with Windows', not 'they are comparing Bash with PowerShell'.
Given that everything else is detailing what about Git Bash is familiar, which as far as I am concerned is everyday business. The author discusses things like Unicode and the other GNU coreutils. Again, this is what you'd expect on a Unix-like.
> I literally just said he wasn’t just comparing Git Bash with Powershell though
That's great that you think that, maybe
delta_p_delta_x had a different takeaway. Why do you believe your opinion about what the cornerstone of the article is to be fact?
> If there’s one thing I can’t stand in IT, it’s people who think their personal preferences (or opinions) are equivalent to impartial facts.
I appreciate what you’re saying but the author literally named the software they were comparing. WSL was specifically mentioned, by name, in some examples.
That’s not subjective opinion of mine because it’s literally printed like that on their site (you can read it for yourself too. It’s the same site for everyone)
In principal I agree with you but if there is one commonly used technology that is actually objectively inferior to the myriad of replacement options and is only still around because its been ubiquitous for decades, its bash.
Architecturally speaking, Powershell (and other modern shells) are a huge improvement. But that’s only half the story.
Syntax wise, I find powershell a step backwards. It’s too verbose. Bash is too far the other way but if I’m using a REPL then I’d rather have something too terse than too verbose. Granted that’s just my preference but we are back to subjective arguments.
Ubiquity is another consideration. If you’re writing multiplatform bootstrapping code then Bash is a better option than Powershell. And we are back to subjective arguments again.
Broad terms are only true if you can agree on a specific context. But then they’re no longer broad and instead you’re relying on everyone agreeing to that narrow subjective definition.
As a thought experiment, Get-ChildItem is a poor choice of API. Instead, it should really just be Get<T> <args>
Get<Folder> -path "C:\Users\Me\My Documents"
Would return a match of type <Folder|nil>
similarly
Find<Folder> -path "C:\Users" -match "Me"
Would return a match of type <Folder[]|nil>
or
Find<Any> -path "C:\Users" -match "Me"
Would return a match of type <Any[]|nil> where Any is
File|Folder|DotFile|Symlink
In as few lines I provided a clean API which unfortunately doesn't exist in any one language but rather borrows from multiple languages, <T> representing C# generic types, File|Folder representing an F# union sum type, Any type a la Rust.
What is better than Bash shell? What I like most about it: It exists on every single Linux box that I use, so I can always run my Bash-specific shell scripts.
> if the writer has been using UNIX shells for that long, why are they using Windows at all?
Desktop applications would probably be the usual reason. It's all well and good if your application needs are covered by open source, but there's a fair bit of software with value add beyond open source that's never received a Linux version and WINE doesn't support well enough.
Of course, as you say, then you have the problem of the shortcomings of various *nix environments for Windows. MacOS can be a great solution... if your app still runs under MacOS (currently I'm trying to figure out how to get Fireworks to run under a minimal windows or mojave VM).
I make a living writing software that's hosted on Windows. I've never felt as comfortable with the command line options on Windows as I do with macos/Linux. Git bash with vim was such a relief.
It's an idiosyncratic mix of ergonomics and habit probably. I don't really care since I get the job done efficiently (more so than a lot of "Windows natives" I observe).
In my experience, it really does come down to the ergonomics: the cursor, the casing, the forward slashes, whatever key combo is required to paste (shift+insert?).
I'm sure PowerShell and friends are great but it's enough of a shift to not be worth the learning curve if it's an environment you're just dipping into to do the bare minimum required to make something work.
PowerShell should be taught in university as example of how poorly designed system devoid of taste and aesthetic looks like. It's an ugly monstrosity that makes me puke every time I try to use it. The only reason some people might like it is because they have been tortured by Windows defaults for many years and finally they got used to this utter mediocracy.
It'd be good if you could actually justify why you think PowerShell is inferior to UNIX shells, instead of going on an impassioned but unsubstantiated rant.
I think powershell is actually really well designed. Some of the syntax leaves a bit to be desired, but other than that I think it's really good. What do you think is poorly designed?
That's quite poor design for a shell! It's like saying an OS is well designed except the mouse and keyboard input leaves a bit to be desired - because they were designed for scripting first and users second.
I'm a back-end developer/DevOps/Linux sysadmin who runs a small consultancy. I've been brought into a few Windows projects -- primarily on-prem/interactive installations. The past few occasions because the client was using Windows because the FE was being built using TouchDesigner. (Not sure it's still Windows-only but it was previously.)
So, it happens. Clients have many reasons for running Windows and depending on the project and its needs, I may accept the requirement or not. To use a recent couple of examples, if I can use GitBash, Docker, WSL and Python locally, then I'll probably take it on. If they're set on using PowerShell, C# and Azure, I'm probably out.
That all being said, if the project is interesting or high profile enough, I'll jump through whatever hoops are put in front of me.
The minor differences can be a major issue. Powershell 7 still doesn't come installed on any windows system as far as I know, so you are better off targeting Powershell 5, and unable to take advantage of the features offered by 7.
I find this interesting because, well, I’ve used bash on Windows for _ages_ (literally before the EMWACs toolkit came out for NT 4.0, and I spent many, many years running Cygwin).
I have to use PowerShell, but since it essentially wraps .NET services I very much prefer to either write C# (and have vastly more maintainable and testable code) or, when removing, invoke the APIs using a Unix scripting language. And these days I can write nice “portable” (ok, retargetable) binaries inside WSL and output a Windows build if I need to…
Git Bash is an elegant solution that brings some of the awesome Gnu utilities and does it on Windows terms and effortlessly. No shoehorn needed.
WSL is handy (those times I'm forced to use Windows - which is still better than when I'm forced to use Mac) but you have to remember the whole time you're in WSL, you're in Linux proper and all things work as Linux, generally, sort of. Git Bash is the power of Gnu right on Windows.
My take here - when working with stable system, changing OS once a several years (last one was when changing laptops basically, so 4+ years in my case), I don't care much on _single_ reboots. For those who tirelessly do distrohopping, things may look different of course.
Years ago, I had a teammate that was always pushing me to try Cygwin. While I love a UNIX shell, Cygwin just felt... weird. The whole paths thing really turned me off. What pushed me to try it? He said to me: Why bother learning two shells (UNIX + DOS), when you can just use one (UNIX) everywhere? I couldn't find a good argument against it. I tried Cygwin, and, eventually (never?) got over the weirdness of paths.
Without question. I was so glad when 22H2 allowed users to change the default console host to Terminal; it was the single biggest reason why I decided to stay on Windows 10.
I don’t respect the attitude of doing whatever you’ve always done, even when working in a wildly different environment.
Imagine how Linux people would react if there were endless blog posts about using VB6 scripts on Linux by running them under WINE because “that’s what feels familiar.”
Sure, for a lark that’s hilarious, but not if you share your work with anyone else for any reason, even open source.
> Imagine how Linux people would react if there were endless blog posts about using VB6 scripts on Linux by running them under WINE because “that’s what feels familiar.”
Heh, that would be hilarious. In fact someone should come up with a Linux distribution that uses NTFS, PowerShell, comes with .NET, WINE, and no coreutils and blog about it.
>Imagine how Linux people would react if there were endless blog posts about using VB6 scripts on Linux by running them under WINE because “that’s what feels familiar.”
Alternatively, we could ask why the POSIX-sphere produces enough power users impassioned to, interested in and capable of enabling their pet projects on a closed and hostile platform for blog posts like the OP's but the reverse does not.
>if the writer has been using UNIX shells for that long, why are they using MSWindows at all?
MSwindows is posix, why not?
windows supports tons of software that is not available on unix. People have to work with other people, spouses, university colleagues, whoever, maybe those people use MSWindows
gitshell is a opinionated msys2 install. they just picked one of the variants and removed the package manager.
this could serve as a wake up call to msys2 maintainers. every time i install it, i have absolutely no f ideia what the 6 variant icons mean and I couldn't care less since all of them mostly work... (thankfully I don't use windows much to reach a point the difference matters?) it's very overdue they just pick one and keep the others with better names in a compatibility folder or something. also, a better name.
If you’re not sure, you probably want UCRT64. At any rate I used it to pull a rabbit out of a hat at work by building some software with it that didn’t have an official windows build.
Building in the URCRT64 environment means the executables wont run on versions of Windows older than 10 without extra runtime support, I think. So that's something to watch out for.
This, corporate has been insisting on using ThreatLocker to block almost everything from running - Git For Windows is the only comfortable environment remaining, at least until I can convince them that this is stupid.
- Paths look much better: C:/Users/doctorpangloss == /Users/doctorpangloss == C:\Users\doctorpangloss.
- The coreutils are just there.
- Works flawlessly as an SSH or Dockerfile shell.
- Fast and tiny. So small you can check it into git.
- The maintainer is very responsive and fixes bugs quickly, even in upstream.
git-bash has problems because of paths, terminal weirdness, and it doesn’t use native Windows APIs for a variety of things where it should. /c/ this, /mnt/ that. It’s not really possible to write multi platform scripts with it. With Busybox for Windows you can.
As a long time personal and professional Linux user, lately my work has led me to work for companies that are Windows-only. I use git bash as my main shell on Windows. It just works for what I need, and in the rare cases I need powershell I just open that.
Most of the windows-based programmers I interact with don't know powershell any better than I do. They use GUI tools for interacting with git and the filesystem.
So for me, git bash for git and filesystem interaction is a superpower in these places.
I could learn powershell, and I'm pretty sure it's much better than bash for scripting, but I'm almost never really scripting in bash anyway, I use tools like F# for that. Powershell seems fine, but it's less well supported on Linux and I don't really need a shell based language.
So in the end, git bash keeps on working and I can focus my learning elsewhere. It's stable, still works the same way 15+ years later, so no need to change.
Always beem 'forced' to use Windows as my workstation by a couple of specialty apps without Linux alternatives that need direct hardware access. But have always used linux on everything else (servers, even my media centre). Because I'm sometimes a late adopter of new tech, until recently I used cygwin as my normal command prompt. There were few downsides to that if you just need access to unix tools, except package management is a bit annoying. Well, discovered the MS Terminal and WSL this year, and happily moved off cygwin. More recently decided to switch over from WS1 to WSL2 but still evaluating the move.
Now I have a unix "more native" environment AND a linux distro I'm familiar with. I've been accessing the Windows filesystem under /mnt for two decades so this is normal for me, and if filesystem access is slower, it's not a dealbreaker. All my existing scripts were trivial to update. The integration of WSL and VSCode is also pretty cool. It means my vs code environment environment is also working natively in my WSL distro.
Overall can understand some lingering negativity about MS efforts in this dept but after years of rejecting all things linux and open source, in general the 180 from the top is much appreciated and it has kept me on Windows a while longer...so mission accomplished microsoft. Just gotta do something about those pesky apps I still need...
git bash's main benefit being that it's a commonly installed variant of the wonderful msys2 system, then yes, I tend to agree.
Another wonderful benefit of git bash, is that because it is so commonly installed, if you want scripting for a dev/build tool, you can 99% of the time just use bash for your scripts, on linux and windows.
I use this all the time for my build scripts. build/x86_64-linux-gnu-gcc.sh, build/x86_64-windows-msvc-cl.sh, etc. (These are generally simple single command line unity builds, parameterized only by debug, opt, etc.)
Then from pwsh I can run these scripts with `& "$env:GIT_INSTALL_ROOT\bin\bash.exe" .\build\x86_64-windows-msvc-cl.sh`, similarly from cmd.exe. (Looks like a ton of typing, but it's not, I always Ctrl-R search and get a hit by just typing "git_").
No need to write a bat script, or even a pwsh one, pretty much ever again. Even if all you want is logic to cd to the script directory, parse a single script argument, check it's valid, provide a default, make some directories if not present, etc., bash beats all the alternatives on many fronts.
I'm "All Linux, all the time" at home, develop in it, help out with the Kernel a little bit, only use Windows at home in a VM, etc.
I'd just started working for a client that is fully entrenched in Windows, but we're doing bare-metal and Linux-y work. When I'd discovered I couldn't use Linux as my daily driver due to security/policy(/tooling) reasons, one of my co-workers told me about Git Bash, and it's worked out so well that ... <looks left, looks right> ... when we were given the opportunity to use Linux but VM Windows, I kept the Windows machine.
GitBash has made Windows painless for my work's use-case. I should probably throw a few $$ at them ....
I have it available, but GitBash seems to work better with my workflow. BTW, when I checked to see how to donate, apparently my client pays for it; I thought it was OOS.
Coming from mac to windows, a big surprise was how ridiculously slow printf is on any of the ms terminals (cmd, powershell, windows terminal). Git bash does not have this problem when you are running something pretty verbose yet want to view the output in the terminal
My favourite shell environment for windows thus far is combining Git For Windows with scoop[1]. A simple "scoop install git" will get the environment installed, and give you a bash shell and full access to all sorts of windows-native utilities from scoop. Some would say I'd be better off with msys2 or cygwin, but the former is meant more as a development environment and lacks misc utilities, and the latter has what is possibly the worst package manager that is still in use (and generally less stellar integration with windows programs).
Scoop favors portable installs and typically manages upgrades and migrations for you. It does support non portable tools but these live in a different bucket.
Chocolatey, and to a greater extent, winget typically defer to the individual applications. In my experience they are more like glorified download indexes than actual repositories for managing software.
Scoop rarely pollutes Windows registry which is what often triggers the corporate blocklists. The portable install is a feature here, compared to chocolatey or winget.
Installing something with chocolatey means downloading some powershell script, with somewhat arbitrary contents, and running that to do all sorts of install, restart, migration, etc., logic, of varying quality, and with varying philosophies of what should be done "automatically".
Installing something with scoop means downloading a declarative json manifest, with occasional, short pre-install/post-install levels of scripting (in all cases I've seen, it amounts to testing if a file is present, and copying it if so, etc.). The complexity that is tolerated for the scripting is much much lower.
I've also found chocolatey to be less well curated, with some odd packages with simple names that come from random people who don't seem to do a great job at maintaining the package.
When I tried using chocolatey, you had to pay them in order to specify a default install directory rather than needing to include it explicitly in every command. I don't know if this is still the case but I formed a deeply held grudge over it.
When I tried using winget, it mostly worked quite well, but occasionally would fail horribly. It'll happily detect and offer to upgrade programs that were installed outside winget, for example - great feature!! - except one time when I tried doing this it installed a new instance to a default location instead of upgrading in place, so now I had two installations. I don't trust it.
Scoop is simple and sensible and just works in exactly the way I expect it to. With the exception of needing to install scoop-search.
winget seems to have replaced chocolatey for the most. Chocolatey is a bit weird with their play to make money, I guess they want to milk it for as much as they can till it dies.
git bash is very cool. It's based on good ol' MSYS2 and MinGW which has always been a nice way to do minimalist Unix-like stuff in Windows.
WSL2 is great these days and can interact fine with Windows files (despite what this article says). But it's a lot more stuff, a full Linux install in a VM.
Tip: Install the IDE inside WSL2. Should also run somewhat faster, especially any operation which has to do with many small files, at the expense of a tiny bit of latency in the GUI.
Really? I always thought the subsystem worked translating OS calls back and forth without virtualizing an entire machine. It sure has an entire OS, kernel and all, but not a virtualization layer. Maybe that was the case for WSL but not WSL2?
WSL2 is "just" Linux running in a Hyper-V VM (with some special sauce on top of it to handle things like interacting with the filesystem or doing Wayland and X11 graphics, plus containerization stuff to allow multiple distributions to be installed and run under one VM and one kernel).
WSL1 was a completely different approach, adding a Linux compatibility layer to Windows itself. There, you never had a Linux kernel running at all-- Linux syscalls would call into WSL, which would talk directly to the NT kernel to do whatever that syscall needed to do.
WSL1 didn't last very long (still present, but not actively being developed)-- turns out that reimplementing one operating system on top of another is a Hard Problem (see also: Wine). WSL2 avoids this entirely, and also avoids most of the impedance mismatches that you get when trying to reimplement POSIX on top of NT. WSL2 solved a whole bunch of compatibility problems essentially overnight that WSL1 never even got around to.
WSL1 used a concept called a pico process [1] and the pico driver that is associated with the process is forwarded the syscalls to translate to the required NT APIs. WSL2 is a VM running through Hyper-V but integrated in a way that mostly looks like a normal process. It was introduced to improve filesystem performance (on the Linux mounts) and avoid having to translate and maintain the syscalls required by Linux [2]. The tech behind WSL1 is quite fun but WSL2 certainly has better compatibility and aside from FS performance between Linux and Windows is mostly a positive.
Yep, WSL1 was a translation layer. Extremely awesome but unfortunately dog slow for anything involving npm packages or large(r) PHP environments as all file accesses would go through the translation layer, the Windows filter chain (aka virus scanners) and then finally NTFS. WSL2 in contrast is a VM, which avoids all the slow paths in file accesses - all that Windows sees is a bunch of bulk block I/O to the backing file of the VM disk.
On the other hand, running Python, Go and .NET on it was a dream. Awesome development environment, and much more battery friendly, to the point where I still use it.
That was WSL1 and as a result had a bunch of unimplemented syscalls.
Microsoft realized trying to keep up with the ever-growing list of syscalls on Linux is actually not practical hence they opted for the virtualized approach with an actual Linux kernel to achieve 1:1 syscall compatibility.
They released that as WSL2. Technically the name is not accurate anymore as there aren’t any “subsystems” at work anymore.
I constantly switch back and forth between Windows and Linux and it's always "dir" mixed up with "ls" coming from my fingertips. Grump grump grump
Edit: I know I can use aliases and .bat files and whatever to make one look like the other, sort of, but then I learn some patchwork system and will be unable to use any other computer's login.
I’ve made the dir and cls aliases on Unix for as long as I can remember. And keep dotfiles in git. Yes, computers out of your control won’t have them, but better to have them 90% (for me 99%) of the time rather than 50%—which is just annoying.
Yes, many windows things accept /, as I’m sure you know. Have used the ssh command instead, easy install now but twenty years ago had to hunt for openssl/ssh for windows. Console/Winterm should allow paste, though console puts on wrong mouse button of course.
> Although I can't get copy/paste to work in a remote text putty window. Sigh.
Depends on stuff like what terminal Putty emulates and so on, of course, but... Eh, have you tried using the old "MS/IBM CUA standard shortcut keys, [Ctrl]-[Ins] and [Shift]-[Ins]? (And, for completeness' sake, [Shift]-[Del] for Cut.)
AIUI, those are Putty's standard keys for that. I've always assumed that's because, you know, a 'Nix command line would interpret [Ctrl]-[C] rather differently.
1. I don’t understand how it relates to msys2 and where the boundary is, so I’m always reluctant to apply msys2 practices (documentation, advice, blog posts, etc.) to Git Bash.
2. I’ve been unable to figure out how to package scripts and other software for Git Bash so other users can install them.
For 2, just tell them to run %GIT_INSTALL_ROOT%\bin\bash.exe yourscript.sh, or you could provide a shim doing exactly that, and install that somewhere in their Path.
Git Bash is really nice for when I want a unix utility that will do the job simpler and better than Powershell, however it's painfully slow for larger I/O operations.
I end up barely using it on my Windows 11 work machine because I'm allowed to have Linux VMs. Even with the VM overhead, and having fewer CPU cores and less RAM than the host, things still end up being way faster there.
I remember one time running a grep command on a large-ish (~ 1 GB) log file in Git Bash and waiting at least 5 minutes for it to complete. After getting impatient I did the same thing in a VM and it took about 30-45 seconds, at which point Git Bash still had not finished.
Did you use endpoint protection in both systems? In my experience, that is often a factor of performance issue on the Windows side (and an unfair comparison when it is not used on Linux)
For those without admin rights preferring ZSH to Bash in the Windows Terminal: install MSYS2 with scoop.sh, then the ZSH package with `pacman -S zsh` and add a shell with commandline `%USERPROFILE%\\scoop\\apps\\msys2\\current\\msys2_shell.cmd -defterm -here -no-start -ucrt64 -shell zsh` to launch it in Windows Terminal.
Oh yeah, I've been using Git Bash for years, it's great. It's nice being able to use a set of basic cli utils I'm used to (gnu utils) which are bundled in (not sure if it's 100% exhaustive in that regard, but still enough for me to do everything familiar). And yes, agree with and have benefited from every single bullet point TFA mentions!
Honestly there are so many differences and little things on Windows that make using any of the CLIs not worth it. You will run up against gotchas developing on native Windows cmd, Powershell, git bash, and WSL. I left for Linux a long time ago and never have had to look back at the utter mess that is Windows. People stay on that platform for what? A graphical menu in the corner? Compatibility you don’t actually need? Use a real shell and OS and save your sanity.
I still love bash and use it as well. But the hate powershell often gets is misguided.