cURL / Mailing Lists / curl-library / Single Mail


libcurl -- the guide

From: Daniel Stenberg <>
Date: Tue, 15 Jan 2002 09:44:05 +0100 (MET)


My initial attempts to write a tutorial-like guide for programming
with libcurl is now found here:

I include the current version here, feel free to comment and improve!



About this Document

 This document will attempt to describe the general principle and some basic
 approach to consider when programming with libcurl. The text will focus
 mainly on the C/C++ interface but might apply fairly well on other interfaces
 as well as they usually follow the C one pretty closely.

 This document will refer to 'the user' as the person writing the source code
 that uses libcurl. That would probably be you or someone in your position.
 What will be generally refered to as 'the program' will be the collected
 source code that you write that is using libcurl for transfers. The program
 is outside libcurl and libcurl is outside of the program.


  Compiling the Program

  Linking the Program with libcurl

  SSL or Not

Global Preparation

 The program must initialize some of the libcurl functionality globally. That
 means it should be done exactly once, no matter how many times you intend to
 use the library. Once for your program's entire life time. This is done using


 and it takes one parameter which is a bit pattern that tells libcurl what to
 intialize. Using CURL_GLOBAL_ALL will make it initialize all known internal
 sub modules, and might be a good default option. The current two bits that
 are specified are:

  CURL_GLOBAL_WIN32 which only does anything on Windows machines. When used on
  a Windows machine, it'll make libcurl intialize the win32 socket
  stuff. Without having that initialized properly, your program cannot use
  sockets properly. You should only do this once for each application, so if
  your program already does this or of another library in use does it, you
  should not tell libcurl to do this as well.

  CURL_GLOBAL_SSL which only does anything on libcurls compiled and built
  SSL-enabled. On these systems, this will make libcurl init OpenSSL properly
  for this application. This is only needed to do once for each application so
  if your program or another library already does this, this bit should not be

 libcurl has a default protection mechanism that detects if curl_global_init()
 hasn't been called by the time curl_easy_perform() is called and if that is
 the case, libcurl runs the function itself with a guessed bit pattern. Please
 note that depending solely on this is not considered nice nor very good.

 When the program no longer uses libcurl, it should call
 curl_global_cleanup(), which is the opposite of the init call. It will then
 do the reversed operations to cleanup the resources the curl_global_init()
 call initialized.

 Repeated calls to curl_global_init() and curl_global_cleanup() should be
 avoided. They should be called once each.

Handle the easy libcurl

 libcurl version 7 is oriented around the so called easy interface. All
 operations in the easy interface are prefixed with 'curl_easy'.

 Future libcurls will also offer the multi interface. More about that
 interface, what it is targeted for and how to use it is still only debated on
 the libcurl mailing list and developer web pages. Join up to discuss and
 figure out!

 To use the easy interface, you must first create yourself an easy handle. You
 need one handle for each easy session you want to perform. Basicly, you
 should use one handle for every thread you plan to use for transferring. You
 must never share the same handle in multiple threads.

 Get an easy handle with

    easyhandle = curl_easy_init();

 It returns an easy handle. Using that you proceed to the next step: setting
 up your preferred actions. A handle is just a logic entity for the upcoming
 transfer or series of transfers. One of the most basic properties to set in
 the handle is the URL. You set your preferred URL to transfer with
 CURLOPT_URL in a manner similar to:

    curl_easy_setopt(easyhandle, CURLOPT_URL, "");

 Let's assume for a while that you want to receive data as the URL indentifies
 a remote resource you want to get here. Since you write a sort of application
 that needs this transfer, I assume that you would like to get the data passed
 to you directly instead of simply getting it passed to stdout. So, you write
 your own function that matches this prototype:

    size_t write_data(void *buffer, size_t size, size_t nmemb, void *userp);

 You tell libcurl to pass all data to this function by issuing a function
 similar to this:

    curl_easy_setopt(easyhandle, CURLOPT_WRITEFUNCTION, write_data);

 You can control what data your function get in the forth argument by setting
 another property:

    curl_easy_setopt(easyhandle, CURLOPT_FILE, &internal_struct);

 Using that property, you can easily pass local data between your application
 and the function that gets invoked by libcurl. libcurl itself won't touch the
 data you pass with CURLOPT_FILE.

 There are of course many more options you can set, and we'll get back to a
 few of them later. Let's instead continue to the actual transfer:

    success = curl_easy_perform(easyhandle);

 The curl_easy_perform() will connect to the remote site, do the necessary
 commands and receive the transfer. Whenever it receives data, it calls the
 callback function we previously set. The function may get one byte at a time,
 or it may get many kilobytes at once. libcurl delivers as much as possible as
 often as possible. Your callback function should return the number of bytes
 it "took care of". If that is not the exact same amount of bytes that was
 passed to it, libcurl will abort the operation and return with an error code.

 When the transfer is complete, the function returns a return code that
 informs you if it succeeded in its mission or not. If a return code isn't
 enough for you, you can use the CURLOPT_ERRORBUFFER to point libcurl to a
 buffer of yours where it'll store a human readable error message as well.

 If you then want to transfer another file, the handle is ready to be used
 again. Mind you, it is even preferred that you re-use an existing handle if
 you intend to make another transfer. libcurl will then attempt to re-use the

When It Doesn't Work

 There will always be times when the transfer fails for some reason. You might
 have set the wrong libcurl option or misunderstood what the libcurl option
 actually does, or the remote server might return non-standard replies that
 confuse the library which then confuses your program.

 There's one golden rule when these things occur: set the CURLOPT_VERBOSE
 option to TRUE. It'll cause the library to spew out the entire protocol
 details it sends, some internal info and some received protcol data as well
 (especially when using FTP). If you're using HTTP, adding the headers in the
 received output to study is also a clever way to get a better understanding
 wht the server behaves the way it does. Include headers in the normal body
 output with CURLOPT_HEADER set TRUE.

Upload Data to a Remote Site

 libcurl tries to keep a protocol independent approach to most transfers, thus
 uploading to a remote FTP site is very similar to uploading data to a HTTP
 server with a PUT request.

 Of course, first you either create an easy handle or you re-use one existing
 one. Then you set the URL to operate on just like before. This is the remote
 URL, that we now will upload.

 Since we write an application, we most likely want libcurl to get the upload
 data by asking us for it. To make it do that, we set the read callback and
 the custom pointer libcurl will pass to our read callback. The read callback
 should have a prototype similar to:

   size_t function(char *buffer, size_t size, size_t nitems, void *userp);

 Where buffer is the pointer to a buffer we fill in with data to upload and
 size*nitems is the size of the buffer. The 'userp' pointer is the custom
 pointer we set to point to a struct of ours to pass private data between the
 application and the callback.

    curl_easy_setopt(easyhandle, CURLOPT_READFUNCTION, read_function);

    curl_easy_setopt(easyhandle, CURLOPT_INFILE, &filedata);

 Tell libcurl that we want to upload:

    curl_easy_setopt(easyhandle, CURLOPT_UPLOAD, TRUE);

 A few protocols won't behave properly when uploads are done without any prior
 knowledge of the expected file size. HTTP PUT is one example [1]. So, set the
 upload file size using the CURLOPT_INFILESIZE like this:

    curl_easy_setopt(easyhandle, CURLOPT_INFILESIZE, file_size);

 So, then you call curl_easy_perform() this time, it'll perform all necessary
 operations and when it has invoked the upload it'll call your supplied
 callback to get the data to upload. The program should return as much data as
 possible in every invoke, as that is likely to make the upload perform as
 fast as possible. The callback should return the number of bytes it wrote in
 the buffer. Returning 0 will signal the end of the upload.


[1] = HTTP PUT without knowing the size prior to transfer is indeed possible,
      but libcurl does not support the chunked transfers on uploading that is
      necessary for this feature to work. We'd gratefully appreciate patches
      that bring this functionality...

    Daniel Stenberg -- curl groks URLs --
Received on 2002-01-15