[dpdk-users] librte_cmdline usage

[Olivier Matz]

Hi,

On Mon, Jan 22, 2018 at 05:15:22PM +0700, longtb5 at viettel.com.vn wrote:
> Hi all,
> 
> I'm writing an application where there already exists a cmdline. Users can
> issue commands by typing from stdin. I want to support another user
> interface where commands are read from sources other than stdin, saved into
> a string buffer, then executed using the already existed cmd_parse_ctx_t.
> 
> So far I have implemented this functionality using the cmdline_interact()
> API like so (pseudo code).
> 
> /* Initialization */
> /* cmds_ctx is the existing command context
>  * that I wish to reuse
>  */
> new_cl = cmdline_new(cmds_ctx, "", cmdbuf_fd, 1); 
> 
> /* Polling */
> for ( ; ; ) {
> 	snprintf(cmdbuf, cmdlen + 2, "%s\n", raw_string);
> 	lseek(cmdbuf_fd, 0 , SEEK_SET);
> 	cmdline_interact(new_cl);
> }
> 
> There are a couple of questions:
> 
> 1. Using cmdline_interact() seems to be quite expensive. How can I reduce
> this cost. Is there a leaner/more elegant way to implement this using DPDK
> API or do I have to write my own code/change the lib?
> 
> 2. How can I disable the behavior where every interaction is written to
> stdout, i.e., to have no output what so ever. Again this is to improve
> performance since writing to stdout is costly.

If my understanding is correct, you want to use the command line parser
without the i/o (readline part). For this, you can directly call
cmdline_parse(cl, buffer). It will parse the given buffer and invokes
the callback.

There will be no output on stdout done by the cmdline library. But it
does not prevent a command callback to call printf() or similar.

Olivier