curl/docs/libcurl/libcurl.3

.\" $Id$
.\"
.TH libcurl 3 "19 March 2002" "libcurl 7.9.6" "libcurl overview"
.SH NAME
libcurl \- client-side URL transfers
.SH DESCRIPTION
This is a short overview on how to use libcurl in your C programs. There are
specific man pages for each function mentioned in here. There are also the
\fIlibcurl-easy(3)\fP man page, the \fIlibcurl-multi(3)\fP man page, the
\fIlibcurl-share(3)\fP man page and the \fIlibcurl-tutorial(3)\fP man page for
in-depth understanding on how to program with libcurl.

There are more than thirty custom bindings available that bring libcurl access
to your favourite language. Look elsewhere for documentation on those.

libcurl has a global constant environment that you must set up and
maintain while using libcurl.  This essentially means you call
\fIcurl_global_init(3)\fP at the start of your program and
\fIcurl_global_cleanup(3)\fP at the end.  See GLOBAL CONSTANTS below
for details.

To transfer files, you always set up an "easy handle" using
\fIcurl_easy_init(3)\fP, but when you want the file(s) transferred you have
the option of using the "easy" interface, or the "multi" interface.

The easy interface is a synchronous interface with which you call
\fIcurl_easy_perform(3)\fP and let it perform the transfer. When it is
completed, the function returns and you can continue. More details are found in
the \fIlibcurl-easy(3)\fP man page.

The multi interface on the other hand is an asynchronous interface, that you
call and that performs only a little piece of the transfer on each invoke. It
is perfect if you want to do things while the transfer is in progress, or
similar. The multi interface allows you to select() on libcurl action, and
even to easily download multiple files simultaneously using a single thread. See further details in the \fIlibcurl-multi(3)\fP man page.

You can have multiple easy handles share certain data, even if they are used
in different threads. This magic is setup using the share interface, as
described in the \fIlibcurl-share(3)\fP man page.

There is also a series of other helpful functions to use, including these:
.RS
.IP curl_version_info()
gets detailed libcurl (and other used libraries) version info
.IP curl_getdate()
converts a date string to time_t
.IP curl_easy_getinfo()
get information about a performed transfer
.IP curl_formadd()
helps building an HTTP form POST
.IP curl_formfree()
free a list built with \fIcurl_formadd(3)\fP
.IP curl_slist_append()
builds a linked list
.IP curl_slist_free_all()
frees a whole curl_slist
.RE

.SH "LINKING WITH LIBCURL"
On unix-like machines, there's a tool named curl-config that gets installed
with the rest of the curl stuff when 'make install' is performed.

curl-config is added to make it easier for applications to link with libcurl
and developers to learn about libcurl and how to use it.

Run 'curl-config --libs' to get the (additional) linker options you need to
link with the particular version of libcurl you've installed. See the
\fIcurl-config(1)\fP man page for further details.

Unix-like operating system that ship libcurl as part of their distributions
often don't provide the curl-config tool, but simply install the library and
headers in the common path for this purpose.

.SH "LIBCURL SYMBOL NAMES"
All public functions in the libcurl interface are prefixed with 'curl_' (with
a lowercase c). You can find other functions in the library source code, but
other prefixes indicate that the functions are private and may change without
further notice in the next release.

Only use documented functions and functionality!
.SH "PORTABILITY"
libcurl works
.B exactly
the same, on any of the platforms it compiles and builds on.
.SH "THREADS"
Never ever call curl-functions simultaneously using the same handle from
several threads. libcurl is thread-safe and can be used in any number of
threads, but you must use separate curl handles if you want to use libcurl in
more than one thread simultaneously.

The global environment functions are not thread-safe.  See GLOBAL CONSTANTS
below for details.

.SH "PERSISTENT CONNECTIONS"
Persistent connections means that libcurl can re-use the same connection for
several transfers, if the conditions are right.

libcurl will \fBalways\fP attempt to use persistent connections. Whenever you
use \fIcurl_easy_perform(3)\fP or \fIcurl_multi_perform(3)\fP, libcurl will
attempt to use an existing connection to do the transfer, and if none exists
it'll open a new one that will be subject for re-use on a possible following
call to \fIcurl_easy_perform(3)\fP or \fIcurl_multi_perform(3)\fP.

To allow libcurl to take full advantage of persistent connections, you should
do as many of your file transfers as possible using the same curl handle. When
you call \fIcurl_easy_cleanup(3)\fP, all the possibly open connections held by
libcurl will be closed and forgotten.

Note that the options set with \fIcurl_easy_setopt(3)\fP will be used on
every repeated \fIcurl_easy_perform(3)\fP call.

.SH "GLOBAL CONSTANTS"
There are a variety of constants that libcurl uses, mainly through its
internal use of other libraries, which are too complicated for the
library loader to set up.  Therefore, a program must call a library
function after the program is loaded and running to finish setting up
the library code.  For example, when libcurl is built for SSL
capability via the GNU TLS library, there is an elaborate tree inside
that library that describes the SSL protocol.

\fIcurl_global_init()\fP is the function that you must call.  This may
allocate resources (e.g. the memory for the GNU TLS tree mentioned
above), so the companion function \fIcurl_global_cleanup()\fP releases
them.

The basic rule for constructing a program that uses libcurl is this:
Call \fIcurl_global_init()\fP, with a \fICURL_GLOBAL_ALL\fP argument,
immediately after the program starts, while it is still only one
thread and before it uses libcurl at all.  Call
\fIcurl_global_cleanup()\fP immediately before the program exits, when
the program is again only one thread and after its last use of
libcurl.

You can call both of these multiple times, as long as all calls meet
these requirements and the number of calls to each is the same.

It isn't actually required that the functions be called at the beginning
and end of the program -- that's just usually the easiest way to do it.
It \fIis\fP required that the functions be called when no other thread
in the program is running.

These global constant functions are \fInot thread safe\fP, so you must
not call them when any other thread in the program is running.  It
isn't good enough that no other thread is using libcurl at the time,
because these functions internally call similar functions of other
libraries, and those functions are similarly thread-unsafe.  You can't
generally know what these libraries are, or whether other threads are
using them.

The global constant situation merits special consideration when the
code you are writing to use libcurl is not the main program, but rather
a modular piece of a program, e.g. another library.  As a module,
your code doesn't know about other parts of the program -- it doesn't
know whether they use libcurl or not.  And its code doesn't necessarily
run at the start and end of the whole program.

A module like this must have global constant functions of its own,
just like \fIcurl_global_init()\fP and \fIcurl_global_cleanup()\fP.
The module thus has control at the beginning and end of the program
and has a place to call the libcurl functions.  Note that if multiple
modules in the program use libcurl, they all will separately call the
libcurl functions, and that's OK because only the first
\fIcurl_global_init()\fP and the last \fIcurl_global_cleanup()\fP in a
program change anything.  (libcurl uses a reference count in static
memory).

In a C++ module, it is common to deal with the global constant
situation by defining a special class that represents the global
constant environment of the module.  A program always has exactly one
object of the class, in static storage.  That way, the program
automatically calls the constructor of the object as the program
starts up and the destructor as it terminates.  As the author of this
libcurl-using module, you can make the constructor call
\fIcurl_global_init()\fP and the destructor call
\fIcurl_global_cleanup()\fP and satisfy libcurl's requirements without
your user having to think about it.

\fIcurl_global_init()\fP has an argument that tells what particular
parts of the global constant environment to set up.  In order to
successfully use any value except \fICURL_GLOBAL_ALL\fP (which says to
set up the whole thing), you must have specific knowledge of internal
workings of libcurl and all other parts of the program of which it is
part.

A special part of the global constant environment is the identity of
the memory allocator.  \fIcurl_global_init()\fP selects the system
default memory allocator, but you can use \fIcurl_global_init_mem()\fP
to supply one of your own.  However, there is no way to use
\fIcurl_global_init_mem()\fP in a modular program -- all modules in
the program that might use libcurl would have to agree on one
allocator.

There is a failsafe in libcurl that makes it usable in simple
situations without you having to worry about the global constant
environment at all: \fIcurl_easy_init()\fP sets up the environment
itself if it hasn't been done yet.  The resources it acquires to do so
get released by the operating system automatically when the program
exits.

This failsafe feature exists mainly for backward compatibility because
there was a time when the global functions didn't exist.  Because it
is sufficient only in the simplest of programs, it is not recommended
for any program to rely on it.