3 ms·

> They also are both designed to glue together independent programs via the pipe abstraction PowerShell and its object pipeline are focused on cmdlets which ar

> They also are both designed to glue together independent programs via the pipe abstraction PowerShell and its object pipeline are focused on cmdlets which are not arbitrary programs, they are .NET classes, and PowerShell objects are .NET objects. Unix shells like sh and bash are focused on tying together separate programs and only have a very minimal set of built-in commands. The focus of each one is very different, despite PowerShell's syntax. Piping objects between cmdlets in PowerShell is analogous to working with objects in Python, Ruby, Perl, etc, not like Unix pipes.

useerup 11y ago
> PowerShell and its object pipeline are focused on cmdlets which are not arbitrary programs, they are .NET classes, and PowerShell objects are .NET objects. Both correct and horribly wrong. PowerShell is indeed focused on cmdlets (and functions) and it's pipeline concept is focused on objects. It is also correct that cmdlets are not arbitrary programs. However, cmdlets are not .NET classes. It is true that you can implement a cmdlet through a .NET class that derives from the base Cmdlet class. But that is just one way to create a cmdlet that can be recognized by PowerShell. You can build cmdlets using PowerShell itself. There is nothing in the PowerShell specification (http://www.microsoft.com/en-us/download/confirmation.aspx?id=36389 http://www.microsoft.com/en-us/download/confirmation.aspx?id...) that require a cmdlet to be a .NET class. If you use PowerShells (rich) introspection function to enquire about a cmdlet, there is nothing objective to give it away as either .NET class based or advanced function based (advanced functions are cmdlets). Yes, you can see if a module (distribution unit of cmdlets, providers, functions etc) are "binary" or "script". That merely reinforces the argument that cmdlets are not classes. Indeed, PowerShell will now dynamically create a new type of cmdlets based on an Open Management Infrastructure (OMI) implementation. Think managed routers, switches, Linux boxes etc. If they expose a REST service that complies with OMI, PowerShell can query the capabilities and create cmdlets in-session for each individual type of equipment. PowerShell has since v3 been able to create WMI/CIM cmdlets. The point is that your assertion that cmdlets are .NET classes is both factually wrong, and grossly misleading. Saying that PowerShell cmdlets are .NET classes is like saying that Yum packages are just archives. Sure, you can pick one apart and say "look! it's a .NET class". I can pick a Yum package apart and say: "look! it's just an archive with some files in it!". That said, it is true that PowerShell builds upon .NET. It is implemented using .NET and the object model is based on .NET. PowerShell's type system is a superset of .NETs - because an administrators tool needs more ad-hoc typing. We have gotten so used to think that sh shells were developed with pipelines like that (streams stdin/stdout) because the commands are like that. In reality, the commands and the (first) shells were developed in tandem so that the commands would fit the shell and vice versa. What Microsoft is doing with PowerShell is challenge that commands must be "like that". If we make the commands "richer" and allow them to use a higher level protocol (objects) for inter-command communication, an even smarter shell can be built. Yes, it does create some friction with traditional commands. PowerShells solution is to regard those as "cmdlets" that just consume sequences of strings (which are objects) and produce sequences of strings. > Unix shells like sh and bash are focused on tying together separate programs and only have a very minimal set of built-in commands. PowerShell is focused on tying together separate cmdlets and only has a very minimal set of build-on commands. See? (Aside: Actually, PowerShell is more terse than any sh shells: It has 0 (zero) "built-in" commands. What it is has is a module concept with auto-discovery. The seemingly "built-in" are not built in. Those cmdlets comes from modules that are distributed alongside the PowerShell binary, and PowerShell auto-loads them as it will any module that is placed in the correct location. sh shells needs commands like e.g. "cd" that manipulate the environment to be built-in. You could never implement "cd" as an external command. In PowerShell "cd" (alias for the Set-Location cmdlet) is distributed through the Microsoft.PowerShell.Management module) > The focus of each one is very different, despite PowerShell's syntax. Piping objects between cmdlets in PowerShell is analogous to working with objects in Python, Ruby, Perl, etc, not like Unix pipes. No it is not. A pipeline is exactly that: A temporary communications channel that is set up for the duration of the (compound) command, and through which multiple objects flow. In PowerShell the individual cmdlets - like commands in a sh pipeline - are active a the same time. From the PowerShell spec: "A pipeline is a series of one or more commands each separated by the pipe operator | (U+007C). Each command receives input from its predecessor and writes output to its successor." Re: Python, Ruby, Perl, etc: You can create a pipeline concept for those languages as well. See for instance pypi (https://code.google.com/p/python-pipelines/ https://code.google.com/p/python-pipelines/). But it is like you want to portrait PowerShell pipelines as mere method invocations. They are considerably more.
- integraton 11y ago
  > your assertion that cmdlets are .NET classes is both factually wrong They are instances of .NET classes that derive from `Cmdlet` and `PSCmdlet` and run in the PowerShell runtime. As stated plain as day in the Microsoft cmdlet documentation: "Cmdlets are instances of .NET Framework classes; they are not stand-alone executables." Your entire comment is an attempt to obfuscate the nature of PowerShell cmdlets, which are actually very simple. >> Piping objects between cmdlets in PowerShell is analogous to working with objects in Python, Ruby, Perl, etc, not like Unix pipes. > No it is not. Yes it is. This is not about the term 'pipeline,' it's about how PowerShell, Unix shells and pipes, and language runtimes work and the scope to which they apply. FACT: PowerShell's object pipeline is internal to the PowerShell / .NET runtime, just as other language features of Python, Ruby, etc are internal to those languages' runtimes. FACT: Unix pipes are primarily for chaining separate processes. They are a lower-level feature than the runtimes of Python, Ruby, or PowerShell, and can be used to chain those runtimes together with each other and other processes. Here's an illustration of the scope of each one. PowerShell's cmdlet / object pipeline: (PowerShell runtime) Unix Pipeline: (Python runtime) | (arbitrary executable) | (PowerShell runtime) | (Ruby runtime)
  - useerup 11y ago
    You look at the developer documentation for developing PowerShell cmdlets using C#, and go "look! it says you create cmdlets using C#!". Well, here is another way to create cmdlets: https://msdn.microsoft.com/en-us/library/jj542520(v=vs.85).aspx https://msdn.microsoft.com/en-us/library/jj542520(v=vs.85).a... Yes, you specify cmdlets using XML. You can point to static methods, remote CIM objects. It may very well use .NET classes underneath. The implementation details are immaterial to how PowerShell works. PowerShell is certainly built using .NET - and require that commands are built (or at least execute) under .NET. However, PowerShell works with objects. An object is a collection of properties and methods that work on the object. Every command in PowerShell must be able to understand the objects, hence a common binary object format must be specified. PowerShell choses .NET objects as it's lingua franca - the PowerShell type system, although it extends it. PowerShell specification says that commands that specifies parameters must do so with types from that type system, and if they support pipelines they must consume objects from the type system and produce objects from the type system. It is a protocol This is very much like how sh shells specifies that commands must support arguments as an array of strings and can optionally consume byte streams from stdin and produce byte streams to stdout. It's a protocol Does PowerShell use the same protocol as sh shells? No. Hence, the commands are different. You do not like that they are different? fine. Don't use it. Your pipeline illustration should have been: PowerShell's cmdlet / object pipeline: (PowerShell runtime | adapter(arbitraty old-school stdin/stdout) | ironpython runtime | any runtime that adheres to type system) Unix Pipeline: (Python runtime) | (arbitrary executable) | (PowerShell runtime) | (Ruby runtime) Yes, PowerShell raises the bar for commands: They must meet some other criteria than string args and stdin/stdout. But just like sh shells and stdin/stdout executables, PowerShell and cmdlets build an ecosystem. It may hurt to realize that the world one knows is just one possible world, and that there may be others out there where some things may be easier and other harder. To me, it is an opportunity to learn. Nothing quite like being shaken up a little.

useerup 11y ago

> PowerShell and its object pipeline are focused on cmdlets which are not arbitrary programs, they are .NET classes, and PowerShell objects are .NET objects. Both correct and horribly wrong. PowerShell is indeed focused on cmdlets (and functions) and it's pipeline concept is focused on objects. It is also correct that cmdlets are not arbitrary programs. However, cmdlets are not .NET classes. It is true that you can implement a cmdlet through a .NET class that derives from the base Cmdlet class. But that is just one way to create a cmdlet that can be recognized by PowerShell. You can build cmdlets using PowerShell itself. There is nothing in the PowerShell specification (http://www.microsoft.com/en-us/download/confirmation.aspx?id=36389 http://www.microsoft.com/en-us/download/confirmation.aspx?id...) that require a cmdlet to be a .NET class. If you use PowerShells (rich) introspection function to enquire about a cmdlet, there is nothing objective to give it away as either .NET class based or advanced function based (advanced functions are cmdlets). Yes, you can see if a module (distribution unit of cmdlets, providers, functions etc) are "binary" or "script". That merely reinforces the argument that cmdlets are not classes. Indeed, PowerShell will now dynamically create a new type of cmdlets based on an Open Management Infrastructure (OMI) implementation. Think managed routers, switches, Linux boxes etc. If they expose a REST service that complies with OMI, PowerShell can query the capabilities and create cmdlets in-session for each individual type of equipment. PowerShell has since v3 been able to create WMI/CIM cmdlets. The point is that your assertion that cmdlets are .NET classes is both factually wrong, and grossly misleading. Saying that PowerShell cmdlets are .NET classes is like saying that Yum packages are just archives. Sure, you can pick one apart and say "look! it's a .NET class". I can pick a Yum package apart and say: "look! it's just an archive with some files in it!". That said, it is true that PowerShell builds upon .NET. It is implemented using .NET and the object model is based on .NET. PowerShell's type system is a superset of .NETs - because an administrators tool needs more ad-hoc typing. We have gotten so used to think that sh shells were developed with pipelines like that (streams stdin/stdout) because the commands are like that. In reality, the commands and the (first) shells were developed in tandem so that the commands would fit the shell and vice versa. What Microsoft is doing with PowerShell is challenge that commands must be "like that". If we make the commands "richer" and allow them to use a higher level protocol (objects) for inter-command communication, an even smarter shell can be built. Yes, it does create some friction with traditional commands. PowerShells solution is to regard those as "cmdlets" that just consume sequences of strings (which are objects) and produce sequences of strings. > Unix shells like sh and bash are focused on tying together separate programs and only have a very minimal set of built-in commands. PowerShell is focused on tying together separate cmdlets and only has a very minimal set of build-on commands. See? (Aside: Actually, PowerShell is more terse than any sh shells: It has 0 (zero) "built-in" commands. What it is has is a module concept with auto-discovery. The seemingly "built-in" are not built in. Those cmdlets comes from modules that are distributed alongside the PowerShell binary, and PowerShell auto-loads them as it will any module that is placed in the correct location. sh shells needs commands like e.g. "cd" that manipulate the environment to be built-in. You could never implement "cd" as an external command. In PowerShell "cd" (alias for the Set-Location cmdlet) is distributed through the Microsoft.PowerShell.Management module) > The focus of each one is very different, despite PowerShell's syntax. Piping objects between cmdlets in PowerShell is analogous to working with objects in Python, Ruby, Perl, etc, not like Unix pipes. No it is not. A pipeline is exactly that: A temporary communications channel that is set up for the duration of the (compound) command, and through which multiple objects flow. In PowerShell the individual cmdlets - like commands in a sh pipeline - are active a the same time. From the PowerShell spec: "A pipeline is a series of one or more commands each separated by the pipe operator | (U+007C). Each command receives input from its predecessor and writes output to its successor." Re: Python, Ruby, Perl, etc: You can create a pipeline concept for those languages as well. See for instance pypi (https://code.google.com/p/python-pipelines/ https://code.google.com/p/python-pipelines/). But it is like you want to portrait PowerShell pipelines as mere method invocations. They are considerably more.

integraton 11y ago
> your assertion that cmdlets are .NET classes is both factually wrong They are instances of .NET classes that derive from `Cmdlet` and `PSCmdlet` and run in the PowerShell runtime. As stated plain as day in the Microsoft cmdlet documentation: "Cmdlets are instances of .NET Framework classes; they are not stand-alone executables." Your entire comment is an attempt to obfuscate the nature of PowerShell cmdlets, which are actually very simple. >> Piping objects between cmdlets in PowerShell is analogous to working with objects in Python, Ruby, Perl, etc, not like Unix pipes. > No it is not. Yes it is. This is not about the term 'pipeline,' it's about how PowerShell, Unix shells and pipes, and language runtimes work and the scope to which they apply. FACT: PowerShell's object pipeline is internal to the PowerShell / .NET runtime, just as other language features of Python, Ruby, etc are internal to those languages' runtimes. FACT: Unix pipes are primarily for chaining separate processes. They are a lower-level feature than the runtimes of Python, Ruby, or PowerShell, and can be used to chain those runtimes together with each other and other processes. Here's an illustration of the scope of each one. PowerShell's cmdlet / object pipeline: (PowerShell runtime) Unix Pipeline: (Python runtime) | (arbitrary executable) | (PowerShell runtime) | (Ruby runtime)
- useerup 11y ago
  You look at the developer documentation for developing PowerShell cmdlets using C#, and go "look! it says you create cmdlets using C#!". Well, here is another way to create cmdlets: https://msdn.microsoft.com/en-us/library/jj542520(v=vs.85).aspx https://msdn.microsoft.com/en-us/library/jj542520(v=vs.85).a... Yes, you specify cmdlets using XML. You can point to static methods, remote CIM objects. It may very well use .NET classes underneath. The implementation details are immaterial to how PowerShell works. PowerShell is certainly built using .NET - and require that commands are built (or at least execute) under .NET. However, PowerShell works with objects. An object is a collection of properties and methods that work on the object. Every command in PowerShell must be able to understand the objects, hence a common binary object format must be specified. PowerShell choses .NET objects as it's lingua franca - the PowerShell type system, although it extends it. PowerShell specification says that commands that specifies parameters must do so with types from that type system, and if they support pipelines they must consume objects from the type system and produce objects from the type system. It is a protocol This is very much like how sh shells specifies that commands must support arguments as an array of strings and can optionally consume byte streams from stdin and produce byte streams to stdout. It's a protocol Does PowerShell use the same protocol as sh shells? No. Hence, the commands are different. You do not like that they are different? fine. Don't use it. Your pipeline illustration should have been: PowerShell's cmdlet / object pipeline: (PowerShell runtime | adapter(arbitraty old-school stdin/stdout) | ironpython runtime | any runtime that adheres to type system) Unix Pipeline: (Python runtime) | (arbitrary executable) | (PowerShell runtime) | (Ruby runtime) Yes, PowerShell raises the bar for commands: They must meet some other criteria than string args and stdin/stdout. But just like sh shells and stdin/stdout executables, PowerShell and cmdlets build an ecosystem. It may hurt to realize that the world one knows is just one possible world, and that there may be others out there where some things may be easier and other harder. To me, it is an opportunity to learn. Nothing quite like being shaken up a little.