I can add that if I use "cat file" on the bash I get the right Greek characters, while the same command via proc:execute returns gibberish characters (the optional argument "utf-8" in proc:execute has no effect). Thanks.

Joseph

Il giorno 09/nov/2015, alle ore 19:04, meumapple meumapple@gmail.com ha scritto:

Hi,

I do not have any LANG variable set. If I redirect into a file the output of the bash command, I get a message with "utf-8" when I analyze this file with the command "file". And indeed the characters are displayed correctly. So the problem arises only when I run everything through proc:execute. If I save the result of the xquery script and analyze them with "file" I get the message "output: ERROR : line 22 : regexec error 17, (illegal byte sequence). And of course the Greek characters are messed up (but not the ASCII ones for English)

Hope this helps...

Joseph

Il giorno 09/nov/2015, alle ore 15:18, Jens Erat jens.erat@uni-konstanz.de ha scritto:

Please report the encoding of your actual files. What is the output of `env | grep LANG` (when stored in your bash script)? If you pipe the output in another file, what will `file -i [output-file]` report (`file` applies some magic to discover file formats and encoding)?

Am 09.11.2015 um 14:06 schrieb meumapple: Hi All,

I have been able to run the command just putting it in a bash file and then running this file with proc:execute. It works. The problem now is with the encoding of the text returned by proc:execute. In my bash I can display Greek characters, but they are messed up if returned by the proc:execute function. I also added utf-8 as a parameter for the function but nothing changes. How can I solve this? Thanks.

Joseph

Il giorno 09/nov/2015, alle ore 10:41, Jens Erat jens.erat@uni-konstanz.de ha scritto:

Hi Joseph,

As far as I understand the documentation, BaseX `proc:system` does not use a shell interpreter by default. `<` or also the pipe `|` are features offered by your shell, not the operating system.

Instead of passing a single parameter, pass the command as parameter to `/usr/bin/env sh -c`, which will make BaseX run a shell that interpretes I/O indirection:

proc:system('/usr/bin/env', ('sh', '-c', 'opennlp postagger model.bin < file.txt'))

This might be a reasonable thing to provide as a built-in function, I'm also not sure whether one can connect some kind of stream/strings to STDIN, STDOUT and STDERR.

Regards from Lake Constance, Germany, Jens

Am 09.11.2015 um 10:19 schrieb Christian Grün:

...

...

Any suggestions?

Any example? ;) Christian

Thanks.

...

Joseph

Inizio messaggio inoltrato:

Da: meumapple meumapple@gmail.com Data: 06 novembre 2015 20:52:08 CET A: BaseX basex-talk@mailman.uni-konstanz.de Oggetto: Bash command

Hi all,

I am using proc:system for a bash command but I cannot run it. It looks like opennlp postagger model.bin < file.txt. I suspect that the problem is the less-than sign, which is converted into &lt; Is there any way to keep the less-than sign and pass it to the bash properly? Thanks.

Joseph