I’m curious about the scenarios that Powershell used to work for you in, and what tooling you’ve moved on to. Do you use something else for Write-Once-Run-(Nearly)Anywhere type scripts?
I primarily write Powershell Core scripts for scenarios where I need to execute the same commands on a variety of operating systems, and I know that the script is likely to be maintained by your “typical” sysadmin (highly technical, but not a programmer)in an environment where installing runtimes for programming languages is discouraged. I switched to macOS as my daily driver about 2 years ago, so PS fits these scenarios pretty well, and Powershell Core updated fairly regularly. Sure there are annoying bugs and misses with the built-in and add-on MS modules: networking cmdlets are an almost total miss, Get-LocalGroup (and maybe other commands?) is totally broken on some AzureAD-joined machines, and the Azure and MgGraph Powershell modules still don’t have enough coverage to move on from the legacy Windows Powershell modules (or even to rely on just one of them, for areas they supposedly cover). But overall I’ve been pretty happy that 99% of the time I can write a powershell script once, and it will run on any machine with Powershell Core, in a consistent way.
Don't do as much "local" sysadmin stuff as most of the workloads moved to the cloud and containers, and that's all automateable via APIs which are much nicer to work with in python with its useful library support. (one of the biggest misses in powershell)
There's still a lot of good stuff wrt powershell maintainability by normal humans (though the entire mental model of object output usually throws them for a loop for years) managing local stuff.
Absolutely, and the calling conventions for all those pieces of code make no sense to powershell folks :)
If you have a background in windows dev or any of those tools then yeah, easy mode go grab your tools, but for most people doing system automation the calling conventions and complete lack of discoverability within the powershell ecosystem (and their tools in no way helping out) made this not a realistic use case.
Shoot I have seen some powershell modules that have to embed string C#s and eval thing just to have basic performance or other basic use cases.
The same challenge with calling conventions happens with Python calling into Windows APIs, moreso since there aren't that many wrappers available, other than installing something like ActiveState Python.
It actually is, as it provides scripting across the whole OS stack and applications, closer to Xerox, Lisp Machines and Smalltalk REPL, something that no UNIX has ever offered, with exception of outliers like NeXTSTEP and OS X.
This is all technically true, but the thesis is that PowerShell makes it all horribly clumsy to use.
From the perspective of the Lisp Machine or, hell, even the AS/400, PowerShell (and, while we're here, the CLR) doesn't quite go deeply enough, pervasively enough, across the system to make it truly useful in the same way.
Extremely clumsy, yes. And not idiomatic - a vanishingly small people learning PS will be able to usefully discover, use, or troubleshoot this utilization.
At the very least they could have helped with the discovery/usage problem, but that would probably have been a really tall order for one little language to do.
Not quite. What I'm more alluding to is something like, say, rewriting smss.exe and csrss.exe to be CLR programs (along with the implication of porting the CLR to the Native API) and giving PowerShell access to their internals directly.
IMO PowerShell is very well integrated with Microsoft Azure cloud. Every Azure resource I work with (Storage, VM, Kusto, EventHub, Service Fabric, AAD, Networking) has tons of PowerShell support. I never set time to use Python with Azure Cloud - and this is only because PowerShell is so good at Azure.
I use Python for AWS, a lot. Last I checked, to use the Azure Python module you have to setup an Enterprise app, which seemed like overkill to me, and not something I think most of our sys admins would want to do. Maybe I’m wrong?
It's a big miss for the casual convenience of a scripter/non-programmer. A Pythonic way to do a remote procedure call is XMLRPC (pseudocode):
import xmlrpc
svr = xmlrpc.connect('http://remote/')
result = svr.add(1, 2)
The C# way is to use the Microsoft Windows Communication Foundation (WFC) Client Proxy using the Service Model Metadata Utility Tool and the Web Services Description Language (WSDL) and XML Schema Definition Language (XSD) files from the remote server, declare a public interface attributed as a Service Contract referencing a namespace, generate a class which inherits from the generic ClientBase<TChannel> and implements the new interface, create an instance of said WCF client and call its methods. (Or rely on Visual Studio magic to hide all that) - https://learn.microsoft.com/en-us/dotnet/framework/wcf/acces...
In any decision, Python goes for "What would Guido do?" and C# gets some union of "what would a committee of Microsoft, IBM, Oracle do?", "What would impress Gartner?", "What is Microsoft legally obliged to do, and backwards-compatibly required to do?", "What would we do if we tried to do everything everyone needs all in one?", "What would Java do?", "What would a large team need to design and maintain a stable, typed, large system for years?".
PowerShell is on top of that; there's no simple included graphing and drawing, no simple hooks into Windows own voice recognition and OCR, and definitely not into whatever magically good ones newer Office / Cloud is using, no casual email or spreadsheet handling, no Visual Basic style form building, no simple data science; there's a few things you can do or download, generally less convenient than a Python equivalent. And Microsoft are leaving it all 'to the community' but the community is using Python so that's where the Excel power-user who wants to script a couple of things will go.
I don’t think Python, a general purpose scripting language; and PowerShell, a shell, are going for the same things. They have different goals and by your definition every other tool is a “big miss” because they are not Python.
I have never desired to do xmlrpc, ocr, voice recognition, or gui building from my shell (and if I did I still don’t see how importing a C# library would be a big miss). What I do desire to do is open files, read their contents and pipe them into other programs, something that Python makes a pain to do with all the file handling. Powershell definitely excels at this, does that make Python a big miss?
I thought I was replying somewhere under this other side-thread https://news.ycombinator.com/item?id=40875610 which says "There are probably users in the millions that are ok at the basics of programming, but don't have the job role to where tools like Java or C# make sense. Python is usually a good fit here, but I really wish Microsoft had something written for us common folks and not just server admins and IT folks. If Microsoft put some more effort into PWSH to where it wasn't turtle slow at things like parsing files and then started adding things like what I talk about above. Maybe even cmdlets for statistics and science...it could be something pretty amazing that your typical business analyst could quickly use"
in that context C# library isn't fit for it. Yes I agree Python and PowerShell aren't going for the same things; that's annoying because PowerShell is 80% of the way there.
> "I have never desired to do ocr, voice recognition, or gui building from my shell"
I have wanted those things. Windows which has a built in speech recognition engine which is COM automatable, a shell (PowerShell) which can be a COM client, and I have a folder full of phonecall recordings. I have a folder full of photos with things like menus and road signs and I want the equivalent of strings.exe for OCR and PowerShell could almost do it. I've wanted to build a Delphi or VB6 or C# style drag-drop GUI and tried to do it in PowerShell with SharpDevelop, WinForms code, the ShowUI module. I've wanted to build a TUI but Windows console host isn't good at those. I've wanted to get jpg metadata out and done it with Shell.Application automation around Explorer instead of downloading mediainfo.
All these thing have something in common - a core in a low level language, glued together or scripted in a high level language. Microsoft have written the core. They have written the high level language. They just didn't bother to make it all integrated for the ordinary power user, or flesh it out with more features along those lines over the years.
> "if I did I still don’t see how importing a C# library would be a big miss"
Because, compared to a builtin "ConvertFrom-Speech" you have to be enough of a programmer to know you need C#, go looking for a package, navigate oneget/winget/psget/nuget/github to download it, worry about .NET version compatibility and module paths, work out how to add an assembly, and then deal with interop, [ref] parameters, byte arrays, streaming.
> "What I do desire to do is open files, read their contents"
Same. And the contents could be all common formats on Windows since the 1990s - MP3, JPG, Excel - things Windows can read and play, things Explorer can read metadata from.
> "and pipe them into other programs, something that Python makes a pain to do with all the file handling. Powershell definitely excels at this, does that make Python a big miss?"
Yes, I think PowerShell is a far more convenient REPL than Python's REPL. Than any REPL I know of, actually - within the boundaries of introspecting small simple data, PowerShell and .NET at least. And yet Python has set()-set() and PowerShell has [system.collections.generic.hashset[psobject]]::new().ExceptWith() (it doesn't return anything it mutates in-place) and every week, people post on the internet asking how to do essentially set union, intersection, subtraction and equality checks in PowerShell and the answer has been unsatisfying for decades.
Adding packages is `dotnet add package {PackageName}`. That's it.
Worrying about version compatibility for new projects has stopped being an issue. The package either targets NS2.0 or whatever latest LTS currently is, in which case you just add its reference, or it doesn't in which case you use something else.
If it does, in 98% situations it just works. In the last 2% it has native dependencies which means either a) the package ships with binaries built for all popular platforms, b) the package adds a platform-specific dependent package automatically, or manually and mentions that in README (either with dotnet add package or system-wide library, apt-get install and friends), or c) the package comes with windows only native dll, which happens with ancient unmaintained packages, it's a rare case nowadays fortunately.
As someone whose primary PL is C#, I found https://github.com/waf/CSharpRepl and https://github.com/dotnet-script/dotnet-script far more accessible and useful. Compilation caching for the latter works relatively well to make startup latency tolerable for using it for writing scripts over Python. It's not the smoothest ride, but the advantages of C# make up for this.
Or I just do `dotnet new console -o MyScriptName --aot`, echo code into Program.cs and `dotnet publish -o .` it. Some do that with Rust as well. Especially useful if you need your script to go through a lot of data quickly and parallelize that well too.
I primarily write Powershell Core scripts for scenarios where I need to execute the same commands on a variety of operating systems, and I know that the script is likely to be maintained by your “typical” sysadmin (highly technical, but not a programmer)in an environment where installing runtimes for programming languages is discouraged. I switched to macOS as my daily driver about 2 years ago, so PS fits these scenarios pretty well, and Powershell Core updated fairly regularly. Sure there are annoying bugs and misses with the built-in and add-on MS modules: networking cmdlets are an almost total miss, Get-LocalGroup (and maybe other commands?) is totally broken on some AzureAD-joined machines, and the Azure and MgGraph Powershell modules still don’t have enough coverage to move on from the legacy Windows Powershell modules (or even to rely on just one of them, for areas they supposedly cover). But overall I’ve been pretty happy that 99% of the time I can write a powershell script once, and it will run on any machine with Powershell Core, in a consistent way.