[ragel-users] Writing a Telnet parser

[Jonathan Castello]

Hi Adrian,

Thanks for your help. Actually, I know how I want to buffer them; the
problem is actually extracting them when I want to. I need some way to
extract the characters only when the next character isn't plain_text
or there is no next character. I've tried adding an entry action to
cr_sequence and iac_sequence, but that doesn't work when you reach the
end of the subject data without seeing a CR or IAC. What I was hoping
to do is maintain a 'left' pointer to the first plain_text character,
and use fpc as the 'right' pointer when I reach the last contiguous
plain_text character. Then I would pass the left pointer and the
length of that contiguous stretch (fpc-left) to the user-provided
callback.

I could copy each character to a temporary buffer, but I was hoping to
avoid extra allocations. I want to just pass pointers into the
original block of text being parsed, so the calling code can do any
copying and allocating required. My entry action attempt was the
closest I could get: it would properly fire before a non plain_text
sequence, but the major issue is that it wouldn't fire at all when it
reached the end of the subject line.

Thanks again,
~Jonathan

On Wed, Sep 29, 2010 at 10:07 AM, Adrian Thurston
<adrian.thurston at esentire.com> wrote:
> Hi Jonathan,
>
> Ragel does not do any buffering of text for you. It's up to you to decide
> how you want to do that, then implement it yourself. There are a couple
> options. You can copy text to a buffer as you move over characters, or you
> can extract them from the input buffer when you need them. The first
> approach is simpler and guaranteed to work without hitches. The second
> technique is faster, but you have to consider buffer block boundaries.
>
> -Adrian
>
> On 10-09-28 08:30 PM, Jonathan Castello wrote:
>>
>> Hello,
>>
>> I'm building a Telnet parser using Ragel, and I'm having an issue
>> making the actions do what I want. I've pasted the machine definition
>> to a gist: http://gist.github.com/602242
>>
>> The issue is a little hard for me to describe, so I'll try to
>> illustrate it as best as I can. If I have a stream of input, and some
>> part of it is "abcdef<IAC><GA>ghi" (where<x>  is a mnemonic for a
>> single byte), I want to emit events as such: text("abcdef"),
>> command("<GA>"), text("ghi"). The caller provides callbacks, and I
>> would pass the data to them as I interpret it.
>>
>> The problem is that I can't figure out how to define actions that
>> would only trigger when the next character doesn't match plain_text
>> (or there's no more data left to parse in that particular packet), so
>> I can get that full stretch of characters. At the moment, I can only
>> get text("a"), text("b"), text("c") etc. to work, i.e. one plain_text
>> match at a time.
>>
>> I suspect the problem is that cr_sequence and iac_sequence are
>> supposed to behave this way - they, too, match singular "terms" each
>> time before returning to the start - but here I am, wanting to give
>> plain_text special treatment. Am I even coming at this from the right
>> angle?
>>
>> Thanks in advance for any advice!
>> ~Jonathan Castello
>>
>> _______________________________________________
>> ragel-users mailing list
>> ragel-users at complang.org
>> http://www.complang.org/mailman/listinfo/ragel-users
>>
>
> _______________________________________________
> ragel-users mailing list
> ragel-users at complang.org
> http://www.complang.org/mailman/listinfo/ragel-users
>

_______________________________________________
ragel-users mailing list
ragel-users at complang.org
http://www.complang.org/mailman/listinfo/ragel-users