Hi! > I have few questions for my pull request: > https://github.com/php/php-src/pull/1217 > > Short story: in some special cases we need to upload just created small > file(s) by CURL. For example it's autogenerated image or pdf. And libcurl > give for us this options: CURLFORM_BUFFER, CURLFORM_BUFFERPTR and > CURLFORM_BUFFERLENGTH. My request about adding this options to curl php > extension.
I wonder if it's be hard to make streams work there... Then you could just pass php://memory or data:// as the file. Not sure about security implications though, needs some thinking. > For basement I used class CURLFile, but can't find solution to use same > class, because this class require file name on disc parameter in > constructor. So I created new class, now it called as CURLBufferFile. While it does require file name as parameter, nobody says anything about disc there. So in addition to the above, you just could add setFileContents() API (or setFileBuffer? feel free to bikeshed :) that would make it take existing buffer instead of reading the file. > So, my questions are: > 1) Can we still integrate this functionality to exists class CURLFile? I think, yes. Or at least we should try to. The meaning of CURLFile is not "file on disk", it's "something to be fed to CURL to be uploaded as a file", so having buffer does not violate the semantics. > 3) What do you think about unserialization for this class? > a) Do not accept it (becouse object can contain very big file) > b) Accept I don't see big issue in serializing/unserializing it. Yes, it can be big. So can be any string you can pass to unserializer, any object can contain tons of data, nothing different here. This object is just container for its values, so I don't see anything special here. -- Stas Malyshev smalys...@gmail.com -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
