Programming Forums > Projects > Existing Project Development

Mad_guy · Mar 15th, 2007, 8:02 PM

So I was trying to figure out exactly what to do with Haskell, as I wanted to experiance the language. I read Don Stewarts excellent "Programming Haskell" 3 part series and it lead me to try and write clones of unix commands in Haskell. So far my program implements roughly 11 commands.

Examples:

(Toggle Plain Text)

[austin@continuum ~]$ hsh -v
haskell hsh v0.4, compiled with ghc-6.6, on linux-i386
[austin@continuum ~]$ hsh cat < dummy.txt
hi
i'm austin
[austin@continuum ~]$ hsh tac < dummy.txt
i'm austin
hi
[austin@continuum ~]$ hsh wc < dummy.txt
2
[austin@continuum ~]$ hsh uname
Linux continuum 2.6.20-ARCH #1 SMP PREEMPT Sat Mar 10 10:12:25 CET 2007 i686
[austin@continuum ~]$ hsh whereis gcc
gcc: /usr/bin/gcc
[austin@continuum ~]$ hsh ls /tmp
...
[austin@continuum ~]$ echo -e "dont mess with me\ndue" | hsh grep "do"
dont mess with me
[austin@continuum ~]$

It also implements other commands such as rmdir, cp, rm, mv, mvdir (had to seperate the two), etc. (grep is my favorite.)

For anybody just wanting the source code:
http://paste.lisp.org/display/37726#4

If anybody would like to try it out, especially on other compilers (which I would appreciate,) you can get a source tarball here:
http://austin.youareinferior.net/scr...0070315.tar.gz (linux/bsd only)

If you try it, tell me what haskell compiler and version you use; I have only tested this on GHC 6.6 (so any hugs98 or nhc98 users are appreciated.) Compilation instructions are included in the README file, all you need is a Compiler and Cabal (I decided to try it for this and absolutely loved it as a build system. I wish more were like it.)

If anybody wants to try it but doesn't have a Haskell compiler, I'm not giving out binaries right now. I need to clean up the option code *a lot* (I figured out Haskell's Getopt, so I'll use that for the option code in the v0.5 release.) Otherwise, you wouldn't probably really know how to work it without looking at the source (at least that's how I feel.) If anybody wants to upload a .tar.gz with it in the topic, that'd be fine as well.

Arevos · Mar 16th, 2007, 3:29 PM

I had some fun pouring through your code, and I hope you won't be offended if I say that I noticed a few areas that look like they could be improved. All those "suceeds" add up to a lot of redundancy, and I wanted to see if Haskell had some mechanism of cutting them down.

I remembered reading an article on Haskell error handling, and I decided that throwDyn and catchDyn looked the way to go. I wrote out some test code, and after solving the most obvious errors, I eventually got:

(Toggle Plain Text)

test.hs:6:5:
    Can't make a derived instance of `Typeable ShellError'
    (You need -fglasgow-exts to derive an instance for this class)
    When deriving instances for type `ShellError'

Adding that extension made it work, but it got a long "hmmm" from me. I decided to investigate why this error handling needed a specific compiler flag set, and eventually came across this article. This gist of it seems to that it some advanced typing mojo for Haskell that is only supported by later versions (so Hugs doesn't support it, but GHC 6.4 and above does). Since GHC supports it, and since I couldn't find any error handling mechanism better than throwDyn/catchDyn, I decided to roll with it.

My first step, then, was to define an error type to handle all potential errors:

(Toggle Plain Text)

data ShellError = UsageError | AppError String String
    deriving (Typeable)

The error can either be a UsageError (the user put in the wrong arguments), or an AppError (there's something wrong with one of our embedded applications). This is essentially just an adaptation of the code in the Haskell error handling article.

Next, I put the ShellError type in the Show type class. This will allow me to use the "show" function on my errors:

(Toggle Plain Text)

usage = "usage: hsh [COMMANDS...] [OPTIONS]"

instance Show ShellError where
    show UsageError = usage
    show (AppError app msg) = printf "%s: %s" app msg

So, if we show the UsageError, we get a little usage message, and if we show the AppError, we get a formatted message (like "ls: can't find dir").

Next, I'll use catchDyn to create a simple wrapper function:

(Toggle Plain Text)

handleErrors m = catchDyn m handler
    where handler :: ShellError -> IO ()
          handler err = hPutStrLn stderr (show err) >>
                        exitWith (ExitFailure 1)

This will run the IO monad passed to it, and if an error occurs, it'll print it to STDERR and exit with failure code 1. I had to explicitly specify the type here, so Haskell knew what error type I wanted to handle.

Armed with handleErrors, I wrapped my main function in it:

(Toggle Plain Text)

main = handleErrors (getArgs >>= parse)

And wrote a quick parse function to test it:

(Toggle Plain Text)

parse []        = throwDyn UsageError
parse ("ls":xs) = ls xs

The UsageError worked fine, so next I decided to see if the AppError would work. Since I figured that I'd be checking for files, directories etc. quite a bit, I create a quick "ensure" function to make my life easier:

(Toggle Plain Text)

ensure :: IO Bool -> ShellError -> IO () -> IO ()
ensure pred err m = do x <- pred
                       if x then m else throwDyn err

This function takes in an error type, a predicate, and some code to run. If the predicate is true, it runs the code, but if it is false, it throws the error instead Thanks to some magic I don't fully understand, when the error is thrown the IO () type is still returned. I'll have to figure out what exactly is going on here, but it's rather neat.

Anyway, all that's left is my ls function:

(Toggle Plain Text)

ls []     = ls ["."]
ls (x:xs) =
    ensure (doesDirectoryExist x) dirError $
        getDirectoryContents x >>= mapM_ putStrLn
    where dirError = AppError "ls" (printf "directory '%s' does not exist" x)

Functionally similar to yours, Mad_guy, but it uses mapM_ to map putStrLn to each file in the directory, instead of using a separate recusive function. It also makes use of the "ensure" function, and the error handling.

Anyway, that's the best I could come up with, but I'm still a Haskell newbie too, so I'm sure this can be improved further. I'm not really that happy with the "ensure" function. It doesn't seem that neat to me.

Here's the full source for reference:

(Toggle Plain Text)

import IO
import System.Exit
import System.Directory
import System.Environment
import Data.Typeable
import Control.Exception
import Text.Printf

usage = "usage: hsh [COMMANDS...] [OPTIONS]"

data ShellError = UsageError | AppError String String
    deriving (Typeable)

instance Show ShellError where
    show UsageError = usage
    show (AppError app msg) = printf "%s: %s" app msg

handleErrors m = catchDyn m handler
    where handler :: ShellError -> IO ()
          handler err = hPutStrLn stderr (show err) >>
                        exitWith (ExitFailure 1)

main = handleErrors (getArgs >>= parse)

parse []        = throwDyn UsageError
parse ("ls":xs) = ls xs

ensure :: IO Bool -> ShellError -> IO () -> IO ()
ensure pred err m = do x <- pred
                       if x then m else throwDyn err

ls []     = ls ["."]
ls (x:xs) =
    ensure (doesDirectoryExist x) dirError $
        getDirectoryContents x >>= mapM_ putStrLn
    where dirError = AppError "ls" (printf "directory '%s' does not exist" x)

Mad_guy · Mar 16th, 2007, 4:10 PM

Thanks for the comments. :>

Last night I was talking to a friend and decided to totally rewrite it. Renaming it to version 1.0 was pretty much obligatory. The new version is substantially different. Instead of having to hard code commands in, the new version is extensible. All you have to do is provide a name that the command will respond to and a callback function. Write your function in a module, and when someone calls say, "hsh ls" it'll execute the LS function. You only have to import your module into a special core module, and add an entry to a list for your command to work (and if you want to distribute your program as source using Cabal's "setup.hs sdist," you must slightly modify the hsh.cabal file, but that's all.) When your command is integrated, configure and build (Cabal script provided, just run 'runhaskell Setup.hs ...')

The usefulness of this is that the main core really doesn't "exist" as far as commands go. All commands take a type [String] and return IO () (I added type signatures to most of the base code, but this shouldn't really be a limitation.) The [String] is the command line arguments passed to your program. This means the commands can be as simple or complex as you want. For example, you could write a new grep implementation that instead of just supporting searching like in my example, it uses the System.Console.GetOpt library and accepts much more complex arguments. This also is a benefit as the core (there're only 4 files in the core, the other ones in the source tree are command modules to be included) is cross platform, so you could write unix/win32 specific commands and modules. The 1.0 distribution, however, comes with some modules that use non-portable libraries, so don't expect this to work on Windows right out of the box (this development was done on my linux system.)

I figured this was the best way to accomplish having an extensible program, just let the users build their commands in if they need them. If Haskell was more easily embeddable in Haskell, I probably would have allows the use of literate haskell source code. I would have followed one of the many tutorials on Scheme in Haskell, but this would really limit what you'd be able to write. So just building modules into the executable seemed like the best code.

The code is still pretty naive and basic, so for maybe a 1.1 release I'll beef up the error handling a lot more. I really just need a README before the 1.0 release is done so you could write a module, but here's the source code, I'll assume you can figure it out (building it is just like building any other cabal application):

http://austin.youareinferior.net/scr...0070316.tar.gz

Mar 16th, 2007, 3:29 PM	#2
Arevos Programming Guru Join Date: Aug 2005 Location: England Posts: 1,499 Rep Power: 4	I had some fun pouring through your code, and I hope you won't be offended if I say that I noticed a few areas that look like they could be improved. All those "suceeds" add up to a lot of redundancy, and I wanted to see if Haskell had some mechanism of cutting them down. I remembered reading an article on Haskell error handling, and I decided that throwDyn and catchDyn looked the way to go. I wrote out some test code, and after solving the most obvious errors, I eventually got: (Toggle Plain Text) test.hs:6:5: Can't make a derived instance of `Typeable ShellError' (You need -fglasgow-exts to derive an instance for this class) When deriving instances for type `ShellError' test.hs:6:5: Can't make a derived instance of `Typeable ShellError' (You need -fglasgow-exts to derive an instance for this class) When deriving instances for type `ShellError' Adding that extension made it work, but it got a long "hmmm" from me. I decided to investigate why this error handling needed a specific compiler flag set, and eventually came across this article. This gist of it seems to that it some advanced typing mojo for Haskell that is only supported by later versions (so Hugs doesn't support it, but GHC 6.4 and above does). Since GHC supports it, and since I couldn't find any error handling mechanism better than throwDyn/catchDyn, I decided to roll with it. My first step, then, was to define an error type to handle all potential errors: (Toggle Plain Text) data ShellError = UsageError \| AppError String String deriving (Typeable) data ShellError = UsageError \| AppError String String deriving (Typeable) The error can either be a UsageError (the user put in the wrong arguments), or an AppError (there's something wrong with one of our embedded applications). This is essentially just an adaptation of the code in the Haskell error handling article. Next, I put the ShellError type in the Show type class. This will allow me to use the "show" function on my errors: (Toggle Plain Text) usage = "usage: hsh [COMMANDS...] [OPTIONS]" instance Show ShellError where show UsageError = usage show (AppError app msg) = printf "%s: %s" app msg usage = "usage: hsh [COMMANDS...] [OPTIONS]" instance Show ShellError where show UsageError = usage show (AppError app msg) = printf "%s: %s" app msg So, if we show the UsageError, we get a little usage message, and if we show the AppError, we get a formatted message (like "ls: can't find dir"). Next, I'll use catchDyn to create a simple wrapper function: (Toggle Plain Text) handleErrors m = catchDyn m handler where handler :: ShellError -> IO () handler err = hPutStrLn stderr (show err) >> exitWith (ExitFailure 1) handleErrors m = catchDyn m handler where handler :: ShellError -> IO () handler err = hPutStrLn stderr (show err) >> exitWith (ExitFailure 1) This will run the IO monad passed to it, and if an error occurs, it'll print it to STDERR and exit with failure code 1. I had to explicitly specify the type here, so Haskell knew what error type I wanted to handle. Armed with handleErrors, I wrapped my main function in it: (Toggle Plain Text) main = handleErrors (getArgs >>= parse) main = handleErrors (getArgs >>= parse) And wrote a quick parse function to test it: (Toggle Plain Text) parse [] = throwDyn UsageError parse ("ls":xs) = ls xs parse [] = throwDyn UsageError parse ("ls":xs) = ls xs The UsageError worked fine, so next I decided to see if the AppError would work. Since I figured that I'd be checking for files, directories etc. quite a bit, I create a quick "ensure" function to make my life easier: (Toggle Plain Text) ensure :: IO Bool -> ShellError -> IO () -> IO () ensure pred err m = do x <- pred if x then m else throwDyn err ensure :: IO Bool -> ShellError -> IO () -> IO () ensure pred err m = do x <- pred if x then m else throwDyn err This function takes in an error type, a predicate, and some code to run. If the predicate is true, it runs the code, but if it is false, it throws the error instead Thanks to some magic I don't fully understand, when the error is thrown the IO () type is still returned. I'll have to figure out what exactly is going on here, but it's rather neat. Anyway, all that's left is my ls function: (Toggle Plain Text) ls [] = ls ["."] ls (x:xs) = ensure (doesDirectoryExist x) dirError $ getDirectoryContents x >>= mapM_ putStrLn where dirError = AppError "ls" (printf "directory '%s' does not exist" x) ls [] = ls ["."] ls (x:xs) = ensure (doesDirectoryExist x) dirError $ getDirectoryContents x >>= mapM_ putStrLn where dirError = AppError "ls" (printf "directory '%s' does not exist" x) Functionally similar to yours, Mad_guy, but it uses mapM_ to map putStrLn to each file in the directory, instead of using a separate recusive function. It also makes use of the "ensure" function, and the error handling. Anyway, that's the best I could come up with, but I'm still a Haskell newbie too, so I'm sure this can be improved further. I'm not really that happy with the "ensure" function. It doesn't seem that neat to me. Here's the full source for reference: (Toggle Plain Text) import IO import System.Exit import System.Directory import System.Environment import Data.Typeable import Control.Exception import Text.Printf usage = "usage: hsh [COMMANDS...] [OPTIONS]" data ShellError = UsageError \| AppError String String deriving (Typeable) instance Show ShellError where show UsageError = usage show (AppError app msg) = printf "%s: %s" app msg handleErrors m = catchDyn m handler where handler :: ShellError -> IO () handler err = hPutStrLn stderr (show err) >> exitWith (ExitFailure 1) main = handleErrors (getArgs >>= parse) parse [] = throwDyn UsageError parse ("ls":xs) = ls xs ensure :: IO Bool -> ShellError -> IO () -> IO () ensure pred err m = do x <- pred if x then m else throwDyn err ls [] = ls ["."] ls (x:xs) = ensure (doesDirectoryExist x) dirError $ getDirectoryContents x >>= mapM_ putStrLn where dirError = AppError "ls" (printf "directory '%s' does not exist" x) import IO import System.Exit import System.Directory import System.Environment import Data.Typeable import Control.Exception import Text.Printf usage = "usage: hsh [COMMANDS...] [OPTIONS]" data ShellError = UsageError \| AppError String String deriving (Typeable) instance Show ShellError where show UsageError = usage show (AppError app msg) = printf "%s: %s" app msg handleErrors m = catchDyn m handler where handler :: ShellError -> IO () handler err = hPutStrLn stderr (show err) >> exitWith (ExitFailure 1) main = handleErrors (getArgs >>= parse) parse [] = throwDyn UsageError parse ("ls":xs) = ls xs ensure :: IO Bool -> ShellError -> IO () -> IO () ensure pred err m = do x <- pred if x then m else throwDyn err ls [] = ls ["."] ls (x:xs) = ensure (doesDirectoryExist x) dirError $ getDirectoryContents x >>= mapM_ putStrLn where dirError = AppError "ls" (printf "directory '%s' does not exist" x)

Mar 16th, 2007, 4:10 PM	#3
Mad_guy Hobbyist Programmer Join Date: Oct 2004 Location: Sandstorm, Techno Club Posts: 238 Rep Power: 4	Thanks for the comments. :> Last night I was talking to a friend and decided to totally rewrite it. Renaming it to version 1.0 was pretty much obligatory. The new version is substantially different. Instead of having to hard code commands in, the new version is extensible. All you have to do is provide a name that the command will respond to and a callback function. Write your function in a module, and when someone calls say, "hsh ls" it'll execute the LS function. You only have to import your module into a special core module, and add an entry to a list for your command to work (and if you want to distribute your program as source using Cabal's "setup.hs sdist," you must slightly modify the hsh.cabal file, but that's all.) When your command is integrated, configure and build (Cabal script provided, just run 'runhaskell Setup.hs ...') The usefulness of this is that the main core really doesn't "exist" as far as commands go. All commands take a type [String] and return IO () (I added type signatures to most of the base code, but this shouldn't really be a limitation.) The [String] is the command line arguments passed to your program. This means the commands can be as simple or complex as you want. For example, you could write a new grep implementation that instead of just supporting searching like in my example, it uses the System.Console.GetOpt library and accepts much more complex arguments. This also is a benefit as the core (there're only 4 files in the core, the other ones in the source tree are command modules to be included) is cross platform, so you could write unix/win32 specific commands and modules. The 1.0 distribution, however, comes with some modules that use non-portable libraries, so don't expect this to work on Windows right out of the box (this development was done on my linux system.) I figured this was the best way to accomplish having an extensible program, just let the users build their commands in if they need them. If Haskell was more easily embeddable in Haskell, I probably would have allows the use of literate haskell source code. I would have followed one of the many tutorials on Scheme in Haskell, but this would really limit what you'd be able to write. So just building modules into the executable seemed like the best code. The code is still pretty naive and basic, so for maybe a 1.1 release I'll beef up the error handling a lot more. I really just need a README before the 1.0 release is done so you could write a module, but here's the source code, I'll assume you can figure it out (building it is just like building any other cabal application): http://austin.youareinferior.net/scr...0070316.tar.gz __________________ happy openBSD and linux user; huhu (Toggle Plain Text) #perl 6 ("hello "~++$i~"!").say while $i < 10; #perl 6 ("hello "~++$i~"!").say while $i < 10;

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Haskell	Arevos	Other Programming Languages	25	Jun 3rd, 2007 4:34 PM
Haskell ideas?	Mad_guy	Other Programming Languages	3	Feb 5th, 2007 9:16 PM

Mar 15th, 2007, 8:02 PM	#1
Mad_guy Hobbyist Programmer Join Date: Oct 2004 Location: Sandstorm, Techno Club Posts: 238 Rep Power: 4	Haskell fun: hsh -- a busybox spinoff in haskell So I was trying to figure out exactly what to do with Haskell, as I wanted to experiance the language. I read Don Stewarts excellent "Programming Haskell" 3 part series and it lead me to try and write clones of unix commands in Haskell. So far my program implements roughly 11 commands. Examples: (Toggle Plain Text) [austin@continuum ~]$ hsh -v haskell hsh v0.4, compiled with ghc-6.6, on linux-i386 [austin@continuum ~]$ hsh cat < dummy.txt hi i'm austin [austin@continuum ~]$ hsh tac < dummy.txt i'm austin hi [austin@continuum ~]$ hsh wc < dummy.txt 2 [austin@continuum ~]$ hsh uname Linux continuum 2.6.20-ARCH #1 SMP PREEMPT Sat Mar 10 10:12:25 CET 2007 i686 [austin@continuum ~]$ hsh whereis gcc gcc: /usr/bin/gcc [austin@continuum ~]$ hsh ls /tmp ... [austin@continuum ~]$ echo -e "dont mess with me\ndue" \| hsh grep "do" dont mess with me [austin@continuum ~]$ [austin@continuum ~]$ hsh -v haskell hsh v0.4, compiled with ghc-6.6, on linux-i386 [austin@continuum ~]$ hsh cat < dummy.txt hi i'm austin [austin@continuum ~]$ hsh tac < dummy.txt i'm austin hi [austin@continuum ~]$ hsh wc < dummy.txt 2 [austin@continuum ~]$ hsh uname Linux continuum 2.6.20-ARCH #1 SMP PREEMPT Sat Mar 10 10:12:25 CET 2007 i686 [austin@continuum ~]$ hsh whereis gcc gcc: /usr/bin/gcc [austin@continuum ~]$ hsh ls /tmp ... [austin@continuum ~]$ echo -e "dont mess with me\ndue" \| hsh grep "do" dont mess with me [austin@continuum ~]$ It also implements other commands such as rmdir, cp, rm, mv, mvdir (had to seperate the two), etc. (grep is my favorite.) For anybody just wanting the source code: http://paste.lisp.org/display/37726#4 If anybody would like to try it out, especially on other compilers (which I would appreciate,) you can get a source tarball here: http://austin.youareinferior.net/scr...0070315.tar.gz (linux/bsd only) If you try it, tell me what haskell compiler and version you use; I have only tested this on GHC 6.6 (so any hugs98 or nhc98 users are appreciated.) Compilation instructions are included in the README file, all you need is a Compiler and Cabal (I decided to try it for this and absolutely loved it as a build system. I wish more were like it.) If anybody wants to try it but doesn't have a Haskell compiler, I'm not giving out binaries right now. I need to clean up the option code a lot (I figured out Haskell's Getopt, so I'll use that for the option code in the v0.5 release.) Otherwise, you wouldn't probably really know how to work it without looking at the source (at least that's how I feel.) If anybody wants to upload a .tar.gz with it in the topic, that'd be fine as well. __________________ happy openBSD and linux user; huhu (Toggle Plain Text) #perl 6 ("hello "~++$i~"!").say while $i < 10; #perl 6 ("hello "~++$i~"!").say while $i < 10;

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)

Programming Forums > Projects > Existing Project Development

Haskell fun: hsh -- a busybox spinoff in haskell