[Swift-devel] New built-in system, request for comments

[Michael Wilde]

As I alluded to in a prior post (but didnt say clearly), Im in favor of 
giving system a varying number of returns, something like:

output = system("command");
(output,error) = system("command");
(output,error,rc) = system("command");

Also letting output and error be either strings or string arrays.

Not sure how much flexibility is best, though.

Regarding treating function failures like app failures: I dont think 
thats good in the case of functions.  Further, in the case of apps, we 
have at least one user request to provide more flexible error handling 
for app failures. So when the user requests an rc (from either), any 
non-zero RC should be returned to the user and not cause the function 
(app or built-in) to be considered failed. Null files can be returned to 
preserve dataflow semantics.  This is not totally clear, and not clearly 
"best", but should should consider it.

Yadu, regarding the current implementation: I was about to send a note 
to swift-user when I decided to test it first.  Im getting strange 
errors when I try to do:

string o[] = system("echo *.swift");
tracef("o=%s\n",o[0]);
# produces:
o=*.swift

string o[] = system("sh -c 'echo *.swift'");
tracef("o=%s\n",o[0]);

# produces:
swift:system returned exitcode :1
swift:system stderr:
  *.swift': -c: line 0: unexpected EOF while looking for matching 
`''*.swift': -c: line 1: syntax error: unexpected end of file

Execution failed:
java.lang.IndexOutOfBoundsException: Invalid index [0] for o
     tracef @ system, line: 28
Caused by: java.lang.IndexOutOfBoundsException: Invalid index [0] for o

---

Yet:
string o[] = system("sh -c 'date'");
tracef("o=%s\n",o[0]);
# produces:
o=Sat Apr  5 20:08:29 CDT 2014

So something strange is happening when I try to pass an argument with 
shell wildcard chars to "sh -c".

I think that system( ) by default should send its args to "sh -c" so 
that wildcards and all related shell features are available.

Lastly, note that in above, "echo" is just an experiment; what I really 
want to do in this example is "/bin/ls -1 *.swift" for example.

- Mike


On 4/5/14, 7:01 PM, Yadu Nand Babuji wrote:
> Ketan,
>
> Do we need that ? If you add stdout, stderr and say exitcode as 
> returns, the user has no
> choice but to define variables to hold those return values.
>
> Mihael mentioned in a separate thread that the behavior should be to 
> fail early in case the
> executed string returns a non-zero exitcode.
>
> -Yadu
>
> On 04/04/2014 10:58 PM, Ketan Maheshwari wrote:
>> Is it possible to handle stdout and/or stderr of system("command") 
>> separately? That way, it might be useful in other places as well.
>>
>>
>> On Fri, Apr 4, 2014 at 12:08 PM, Yadu Nand <yadudoc1729 at gmail.com 
>> <mailto:yadudoc1729 at gmail.com>> wrote:
>>
>>     Hi,
>>
>>     I have just added a new builtin "system" to trunk. It takes a string,
>>     which is executed in the shell
>>     by java and the results are returned as an array. This is done to
>>     make
>>     mappings easier, for example,
>>     you can map an entire folder "foo" using the following :
>>
>>     file folder[ ] < array_mapper; files = system ("find foo -type f") >;
>>
>>     If the execution of the string provided fails, the logger would
>>     report
>>     the the exitcode and stderr to
>>     the swift stdout.
>>
>>     Thanks,
>>     Yadu Nand B
>>     _______________________________________________
>>     Swift-devel mailing list
>>     Swift-devel at ci.uchicago.edu <mailto:Swift-devel at ci.uchicago.edu>
>>     https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-devel
>>
>>
>>
>>
>> _______________________________________________
>> Swift-devel mailing list
>> Swift-devel at ci.uchicago.edu
>> https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-devel
>
>
>
> _______________________________________________
> Swift-devel mailing list
> Swift-devel at ci.uchicago.edu
> https://lists.ci.uchicago.edu/cgi-bin/mailman/listinfo/swift-devel

-- 
Michael Wilde
Mathematics and Computer Science          Computation Institute
Argonne National Laboratory               The University of Chicago

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/swift-devel/attachments/20140405/14b62fc0/attachment.html>