Re: Apache 2.0 sapi module

On Wed, 25 Oct 2000, Rasmus Lerdorf wrote: > Should be a php-dev msg, but the silly server is still down, so I will > start this here for now. > > You may have noticed that I checked in an initial framework for getting > PHP to work as a filter. The main sticky point right now is that when PHP > is acting like a filter we get a trickle of buckets coming down the filter > chain. What I have currently done is write a bit of code that collects > all the buckets until we have the entire stream of text to be parsed in a > char *. We now need an efficient way to push this stream of data into the > Zend engine to be parsed. I looked at the eval() code, but I am a bit > uncertain about the efficiency of that approach. > > A different approach might be to modify the yyinput stuff to read the > buckets directly one by one. Alternatively, generalize it to a point where you can pass a function reference to the engine: size_t fill_input_buffer(char **); This interface would allow us to keep the real logic (handling buckets) where it belongs--in the Apache 2 SAPI module, while granting SAPI modules enough flexibility to not force them to write out scripts to files first. I'm not much of a Bison expert to think about the possibility/feasibility of such an implementation, so I'd appreciate your comments. - Sascha