luna - curl/docs/libcurl/libcurl-tutorial.md

   1---
   2c: Copyright (C) Daniel Stenberg, <daniel@haxx.se>, et al.
   3SPDX-License-Identifier: curl
   4Title: libcurl-tutorial
   5Section: 3
   6Source: libcurl
   7See-also:
   8  - libcurl-easy (3)
   9  - libcurl-errors (3)
  10  - libcurl-multi (3)
  11  - libcurl-url (3)
  12Protocol:
  13  - All
  14Added-in: n/a
  15---
  16
  17# NAME
  18
  19libcurl-tutorial - libcurl programming tutorial
  20
  21# Objective
  22
  23This document attempts to describe the general principles and some basic
  24approaches to consider when programming with libcurl. The text focuses on the
  25C interface but should apply fairly well on other language bindings as well as
  26they usually follow the C API pretty closely.
  27
  28This document refers to 'the user' as the person writing the source code that
  29uses libcurl. That would probably be you or someone in your position. What is
  30generally referred to as 'the program' is the collected source code that you
  31write that is using libcurl for transfers. The program is outside libcurl and
  32libcurl is outside of the program.
  33
  34To get more details on all options and functions described herein, please
  35refer to their respective man pages.
  36
  37# Building
  38
  39There are many different ways to build C programs. This chapter assumes a Unix
  40style build process. If you use a different build system, you can still read
  41this to get general information that may apply to your environment as well.
  42
  43## Compiling the Program
  44
  45Your compiler needs to know where the libcurl headers are located. Therefore
  46you must set your compiler's include path to point to the directory where you
  47installed them. The 'curl-config'[3] tool can be used to get this information:
  48~~~c
  49    $ curl-config --cflags
  50~~~
  51
  52## Linking the Program with libcurl
  53
  54When having compiled the program, you need to link your object files to create
  55a single executable. For that to succeed, you need to link with libcurl and
  56possibly also with other libraries that libcurl itself depends on. Like the
  57OpenSSL libraries, but even some standard OS libraries may be needed on the
  58command line. To figure out which flags to use, once again the 'curl-config'
  59tool comes to the rescue:
  60~~~c
  61    $ curl-config --libs
  62~~~
  63
  64## SSL or Not
  65
  66libcurl can be built and customized in many ways. One of the things that
  67varies from different libraries and builds is the support for SSL-based
  68transfers, like HTTPS and FTPS. If a supported SSL library was detected
  69properly at build-time, libcurl is built with SSL support. To figure out if an
  70installed libcurl has been built with SSL support enabled, use *curl-config*
  71like this:
  72
  73~~~c
  74    $ curl-config --feature
  75~~~
  76
  77If SSL is supported, the keyword *SSL* is written to stdout, possibly together
  78with a other features that could be either on or off on for different
  79libcurls.
  80
  81See also the "Features libcurl Provides" further down.
  82
  83## autoconf macro
  84
  85When you write your configure script to detect libcurl and setup variables
  86accordingly, we offer a macro that probably does everything you need in this
  87area. See docs/libcurl/libcurl.m4 file - it includes docs on how to use it.
  88
  89# Portable Code in a Portable World
  90
  91The people behind libcurl have put a considerable effort to make libcurl work
  92on a large amount of different operating systems and environments.
  93
  94You program libcurl the same way on all platforms that libcurl runs on. There
  95are only a few minor details that differ. If you make sure to write your code
  96portable enough, you can create a portable program. libcurl should not stop
  97you from that.
  98
  99# Global Preparation
 100
 101The program must initialize some of the libcurl functionality globally. That
 102means it should be done exactly once, no matter how many times you intend to
 103use the library. Once for your program's entire life time. This is done using
 104~~~c
 105    curl_global_init()
 106~~~
 107and it takes one parameter which is a bit pattern that tells libcurl what to
 108initialize. Using *CURL_GLOBAL_ALL* makes it initialize all known internal
 109sub modules, and might be a good default option. The current two bits that are
 110specified are:
 111
 112## CURL_GLOBAL_WIN32
 113
 114which only does anything on Windows machines. When used on a Windows machine,
 115it makes libcurl initialize the Win32 socket stuff. Without having that
 116initialized properly, your program cannot use sockets properly. You should
 117only do this once for each application, so if your program already does this
 118or of another library in use does it, you should not tell libcurl to do this
 119as well.
 120
 121## CURL_GLOBAL_SSL
 122
 123which only does anything on libcurls compiled and built SSL-enabled. On these
 124systems, this makes libcurl initialize the SSL library properly for this
 125application. This only needs to be done once for each application so if your
 126program or another library already does this, this bit should not be needed.
 127
 128libcurl has a default protection mechanism that detects if
 129curl_global_init(3) has not been called by the time
 130curl_easy_perform(3) is called and if that is the case, libcurl runs the
 131function itself with a guessed bit pattern. Please note that depending solely
 132on this is not considered nice nor good.
 133
 134When the program no longer uses libcurl, it should call
 135curl_global_cleanup(3), which is the opposite of the init call. It
 136performs the reversed operations to cleanup the resources the
 137curl_global_init(3) call initialized.
 138
 139Repeated calls to curl_global_init(3) and curl_global_cleanup(3)
 140should be avoided. They should only be called once each.
 141
 142# Features libcurl Provides
 143
 144It is considered best-practice to determine libcurl features at runtime rather
 145than at build-time (if possible of course). By calling
 146curl_version_info(3) and checking out the details of the returned
 147struct, your program can figure out exactly what the currently running libcurl
 148supports.
 149
 150# Two Interfaces
 151
 152libcurl first introduced the so called easy interface. All operations in the
 153easy interface are prefixed with 'curl_easy'. The easy interface lets you do
 154single transfers with a synchronous and blocking function call.
 155
 156libcurl also offers another interface that allows multiple simultaneous
 157transfers in a single thread, the so called multi interface. More about that
 158interface is detailed in a separate chapter further down. You still need to
 159understand the easy interface first, so please continue reading for better
 160understanding.
 161
 162# Handle the Easy libcurl
 163
 164To use the easy interface, you must first create yourself an easy handle. You
 165need one handle for each easy session you want to perform. You should use one
 166handle for every thread you plan to use for transferring. You must never share
 167the same handle in multiple threads.
 168
 169Get an easy handle with
 170~~~c
 171    handle = curl_easy_init();
 172~~~
 173It returns an easy handle. Using that you proceed to the next step: setting
 174up your preferred actions. A handle is a logic entity for the upcoming
 175transfer or series of transfers.
 176
 177You set properties and options for this handle using
 178curl_easy_setopt(3). They control how the subsequent transfer or
 179transfers using this handle are made. Options remain set in the handle until
 180set again to something different. They are sticky. Multiple requests using the
 181same handle use the same options.
 182
 183If you at any point would like to blank all previously set options for a
 184single easy handle, you can call curl_easy_reset(3) and you can also
 185make a clone of an easy handle (with all its set options) using
 186curl_easy_duphandle(3).
 187
 188Many of the options you set in libcurl are "strings", pointers to data
 189terminated with a zero byte. When you set strings with
 190curl_easy_setopt(3), libcurl makes its own copy so that they do not need
 191to be kept around in your application after being set[4].
 192
 193One of the most basic properties to set in the handle is the URL. You set your
 194preferred URL to transfer with CURLOPT_URL(3) in a manner similar to:
 195
 196~~~c
 197    curl_easy_setopt(handle, CURLOPT_URL, "https://example.com/");
 198~~~
 199
 200Let's assume for a while that you want to receive data as the URL identifies a
 201remote resource you want to get here. Since you write a sort of application
 202that needs this transfer, I assume that you would like to get the data passed
 203to you directly instead of getting it passed to stdout. You write your
 204own function that matches this prototype:
 205~~~c
 206    size_t write_data(char *buffer, size_t size, size_t nmemb, void *userp);
 207~~~
 208
 209You tell libcurl to pass all data to this function by issuing a function
 210similar to this:
 211
 212~~~c
 213    curl_easy_setopt(handle, CURLOPT_WRITEFUNCTION, write_data);
 214~~~
 215
 216You can control what data your callback function gets in the fourth argument
 217by setting another property:
 218
 219~~~c
 220    curl_easy_setopt(handle, CURLOPT_WRITEDATA, &internal_struct);
 221~~~
 222Using that property, you can easily pass local data between your application
 223and the function that gets invoked by libcurl. libcurl itself does not touch
 224the data you pass with CURLOPT_WRITEDATA(3).
 225
 226libcurl offers its own default internal callback that takes care of the data
 227if you do not set the callback with CURLOPT_WRITEFUNCTION(3). It
 228outputs the received data to stdout. You can have the default callback write
 229the data to a different file handle by passing a 'FILE *' to a file opened for
 230writing with the CURLOPT_WRITEDATA(3) option.
 231
 232Now, we need to take a step back and take a deep breath. Here is one of those
 233rare platform-dependent nitpicks. Did you spot it? On some platforms[2],
 234libcurl is not able to operate on file handles opened by the
 235program. Therefore, if you use the default callback and pass in an open file
 236handle with CURLOPT_WRITEDATA(3), libcurl crashes. You should avoid this
 237to make your program run fine virtually everywhere.
 238
 239(CURLOPT_WRITEDATA(3) was formerly known as *CURLOPT_FILE*. Both names still
 240work and do the same thing).
 241
 242If you are using libcurl as a Windows DLL, you MUST use the
 243CURLOPT_WRITEFUNCTION(3) if you set CURLOPT_WRITEDATA(3) - or experience
 244crashes.
 245
 246There are of course many more options you can set, and we get back to a few of
 247them later. Let's instead continue to the actual transfer:
 248
 249~~~c
 250    success = curl_easy_perform(handle);
 251~~~
 252
 253curl_easy_perform(3) connects to the remote site, does the necessary commands
 254and performs the transfer. Whenever it receives data, it calls the callback
 255function we previously set. The function may get one byte at a time, or it may
 256get many kilobytes at once. libcurl delivers as much as possible as often as
 257possible. Your callback function should return the number of bytes it "took
 258care of". If that is not the same amount of bytes that was passed to it,
 259libcurl aborts the operation and returns with an error code.
 260
 261When the transfer is complete, the function returns a return code that informs
 262you if it succeeded in its mission or not. If a return code is not enough for
 263you, you can use the CURLOPT_ERRORBUFFER(3) to point libcurl to a buffer of
 264yours where it stores a human readable error message as well.
 265
 266If you then want to transfer another file, the handle is ready to be used
 267again. It is even preferred and encouraged that you reuse an existing handle
 268if you intend to make another transfer. libcurl then attempts to reuse a
 269previous connection.
 270
 271For some protocols, downloading a file can involve a complicated process of
 272logging in, setting the transfer mode, changing the current directory and
 273finally transferring the file data. libcurl takes care of all that
 274complication for you. Given the URL to a file, libcurl takes care of
 275all the details needed to get the file moved from one machine to another.
 276
 277# Multi-threading Issues
 278
 279libcurl is thread-safe but there are a few exceptions. Refer to
 280libcurl-thread(3) for more information.
 281
 282# When It does not Work
 283
 284There are times when the transfer fails for some reason. You might have set
 285the wrong libcurl option or misunderstood what the libcurl option actually
 286does, or the remote server might return non-standard replies that confuse the
 287library which then confuses your program.
 288
 289There is one golden rule when these things occur: set the
 290CURLOPT_VERBOSE(3) option to 1. it causes the library to spew out the
 291entire protocol details it sends, some internal info and some received
 292protocol data as well (especially when using FTP). If you are using HTTP,
 293adding the headers in the received output to study is also a clever way to get
 294a better understanding why the server behaves the way it does. Include headers
 295in the normal body output with CURLOPT_HEADER(3) set 1.
 296
 297Of course, there are bugs left. We need to know about them to be able to fix
 298them, so we are dependent on your bug reports. When you do report suspected
 299bugs in libcurl, please include as many details as you possibly can: a protocol
 300dump that CURLOPT_VERBOSE(3) produces, library version, as much as possible of
 301your code that uses libcurl, operating system name and version, compiler name
 302and version etc.
 303
 304If CURLOPT_VERBOSE(3) is not enough, you increase the level of debug
 305data your application receive by using the CURLOPT_DEBUGFUNCTION(3).
 306
 307Getting some in-depth knowledge about the protocols involved is never wrong,
 308and if you are trying to do funny things, you might understand libcurl and how
 309to use it better if you study the appropriate RFC documents at least briefly.
 310
 311# Upload Data to a Remote Site
 312
 313libcurl tries to keep a protocol independent approach to most transfers, thus
 314uploading to a remote FTP site is similar to uploading data to an HTTP server
 315with a PUT request.
 316
 317Of course, first you either create an easy handle or you reuse one existing
 318one. Then you set the URL to operate on like before. This is the remote URL,
 319that we now upload.
 320
 321Since we write an application, we most likely want libcurl to get the upload
 322data by asking us for it. To make it do that, we set the read callback and the
 323custom pointer libcurl passes to our read callback. The read callback should
 324have a prototype similar to:
 325~~~c
 326    size_t function(char *bufptr, size_t size, size_t nitems, void *userp);
 327~~~
 328Where *bufptr* is the pointer to a buffer we fill in with data to upload
 329and *nitems* is the size of the buffer and therefore also the maximum
 330amount of data we can return to libcurl in this call. The *userp* pointer
 331is the custom pointer we set to point to a struct of ours to pass private data
 332between the application and the callback.
 333~~~c
 334    curl_easy_setopt(handle, CURLOPT_READFUNCTION, read_function);
 335
 336    curl_easy_setopt(handle, CURLOPT_READDATA, &filedata);
 337~~~
 338Tell libcurl that we want to upload:
 339~~~c
 340    curl_easy_setopt(handle, CURLOPT_UPLOAD, 1L);
 341~~~
 342A few protocols do not behave properly when uploads are done without any prior
 343knowledge of the expected file size. Set the upload file size using the
 344CURLOPT_INFILESIZE_LARGE(3) for all known file sizes like this[1]:
 345
 346~~~c
 347    /* in this example, file_size must be an curl_off_t variable */
 348    curl_easy_setopt(handle, CURLOPT_INFILESIZE_LARGE, file_size);
 349~~~
 350
 351When you call curl_easy_perform(3) this time, it performs all the
 352necessary operations and when it has invoked the upload it calls your supplied
 353callback to get the data to upload. The program should return as much data as
 354possible in every invoke, as that is likely to make the upload perform as fast
 355as possible. The callback should return the number of bytes it wrote in the
 356buffer. Returning 0 signals the end of the upload.
 357
 358# Passwords
 359
 360Many protocols use or even require that username and password are provided
 361to be able to download or upload the data of your choice. libcurl offers
 362several ways to specify them.
 363
 364Most protocols support that you specify the name and password in the URL
 365itself. libcurl detects this and use them accordingly. This is written like
 366this:
 367~~~c
 368    protocol://user:password@example.com/path/
 369~~~
 370If you need any odd letters in your username or password, you should enter
 371them URL encoded, as %XX where XX is a two-digit hexadecimal number.
 372
 373libcurl also provides options to set various passwords. The username and
 374password as shown embedded in the URL can instead get set with the
 375CURLOPT_USERPWD(3) option. The argument passed to libcurl should be a
 376char * to a string in the format "user:password". In a manner like this:
 377
 378~~~c
 379    curl_easy_setopt(handle, CURLOPT_USERPWD, "myname:thesecret");
 380~~~
 381
 382Another case where name and password might be needed at times, is for those
 383users who need to authenticate themselves to a proxy they use. libcurl offers
 384another option for this, the CURLOPT_PROXYUSERPWD(3). Its use is similar to the
 385CURLOPT_USERPWD(3) option, like this:
 386
 387~~~c
 388    curl_easy_setopt(handle, CURLOPT_PROXYUSERPWD, "myname:thesecret");
 389~~~
 390
 391There is a long time Unix "standard" way of storing FTP usernames and
 392passwords, namely in the $HOME/.netrc file (on Windows, libcurl also checks
 393the *%USERPROFILE% environment* variable if *%HOME%* is unset, and tries
 394"_netrc" as name). The file should be made private so that only the user may
 395read it (see also the "Security Considerations" chapter), as it might contain
 396the password in plain text. libcurl has the ability to use this file to figure
 397out what set of username and password to use for a particular host. As an
 398extension to the normal functionality, libcurl also supports this file for
 399non-FTP protocols such as HTTP. To make curl use this file, use the
 400CURLOPT_NETRC(3) option:
 401
 402~~~c
 403    curl_easy_setopt(handle, CURLOPT_NETRC, 1L);
 404~~~
 405
 406A basic example of how such a .netrc file may look like:
 407
 408~~~c
 409    machine myhost.mydomain.com
 410    login userlogin
 411    password secretword
 412~~~
 413
 414All these examples have been cases where the password has been optional, or
 415at least you could leave it out and have libcurl attempt to do its job
 416without it. There are times when the password is not optional, like when
 417you are using an SSL private key for secure transfers.
 418
 419To pass the known private key password to libcurl:
 420~~~c
 421    curl_easy_setopt(handle, CURLOPT_KEYPASSWD, "keypassword");
 422~~~
 423
 424# HTTP Authentication
 425
 426The previous chapter showed how to set username and password for getting URLs
 427that require authentication. When using the HTTP protocol, there are many
 428different ways a client can provide those credentials to the server and you
 429can control which way libcurl uses them. The default HTTP authentication
 430method is called 'Basic', which is sending the name and password in clear-text
 431in the HTTP request, base64-encoded. This is insecure.
 432
 433At the time of this writing, libcurl can be built to use: Basic, Digest, NTLM,
 434Negotiate (SPNEGO). You can tell libcurl which one to use with
 435CURLOPT_HTTPAUTH(3) as in:
 436
 437~~~c
 438    curl_easy_setopt(handle, CURLOPT_HTTPAUTH, CURLAUTH_DIGEST);
 439~~~
 440
 441When you send authentication to a proxy, you can also set authentication type
 442the same way but instead with CURLOPT_PROXYAUTH(3):
 443
 444~~~c
 445    curl_easy_setopt(handle, CURLOPT_PROXYAUTH, CURLAUTH_NTLM);
 446~~~
 447
 448Both these options allow you to set multiple types (by ORing them together),
 449to make libcurl pick the most secure one out of the types the server/proxy
 450claims to support. This method does however add a round-trip since libcurl
 451must first ask the server what it supports:
 452
 453~~~c
 454    curl_easy_setopt(handle, CURLOPT_HTTPAUTH, CURLAUTH_DIGEST | CURLAUTH_BASIC);
 455~~~
 456
 457For convenience, you can use the *CURLAUTH_ANY* define (instead of a list with
 458specific types) which allows libcurl to use whatever method it wants.
 459
 460When asking for multiple types, libcurl picks the available one it considers
 461"best" in its own internal order of preference.
 462
 463# HTTP POSTing
 464
 465We get many questions regarding how to issue HTTP POSTs with libcurl the
 466proper way. This chapter thus includes examples using both different versions
 467of HTTP POST that libcurl supports.
 468
 469The first version is the simple POST, the most common version, that most HTML
 470pages using the \<form\> tag uses. We provide a pointer to the data and tell
 471libcurl to post it all to the remote site:
 472
 473~~~c
 474    char *data = "name=daniel&project=curl";
 475    curl_easy_setopt(handle, CURLOPT_POSTFIELDS, data);
 476    curl_easy_setopt(handle, CURLOPT_URL, "https://posthere.example/");
 477
 478    curl_easy_perform(handle); /* post away! */
 479~~~
 480
 481Simple enough, huh? Since you set the POST options with the
 482CURLOPT_POSTFIELDS(3), this automatically switches the handle to use
 483POST in the upcoming request.
 484
 485What if you want to post binary data that also requires you to set the
 486Content-Type: header of the post? Well, binary posts prevent libcurl from being
 487able to do strlen() on the data to figure out the size, so therefore we must
 488tell libcurl the size of the post data. Setting headers in libcurl requests are
 489done in a generic way, by building a list of our own headers and then passing
 490that list to libcurl.
 491
 492~~~c
 493    struct curl_slist *headers = NULL;
 494    headers = curl_slist_append(headers, "Content-Type: text/xml");
 495
 496    /* post binary data */
 497    curl_easy_setopt(handle, CURLOPT_POSTFIELDS, binaryptr);
 498
 499    /* set the size of the postfields data */
 500    curl_easy_setopt(handle, CURLOPT_POSTFIELDSIZE, 23L);
 501
 502    /* pass our list of custom made headers */
 503    curl_easy_setopt(handle, CURLOPT_HTTPHEADER, headers);
 504
 505    curl_easy_perform(handle); /* post away! */
 506
 507    curl_slist_free_all(headers); /* free the header list */
 508~~~
 509
 510While the simple examples above cover the majority of all cases where HTTP
 511POST operations are required, they do not do multi-part formposts. Multi-part
 512formposts were introduced as a better way to post (possibly large) binary data
 513and were first documented in the RFC 1867 (updated in RFC 2388). They are
 514called multi-part because they are built by a chain of parts, each part being
 515a single unit of data. Each part has its own name and contents. You can in
 516fact create and post a multi-part formpost with the regular libcurl POST
 517support described above, but that would require that you build a formpost
 518yourself and provide to libcurl.
 519
 520To make that easier, libcurl provides a MIME API consisting in several
 521functions: using those, you can create and fill a multi-part form. Function
 522curl_mime_init(3) creates a multi-part body; you can then append new parts
 523to a multi-part body using curl_mime_addpart(3).
 524
 525There are three possible data sources for a part: memory using
 526curl_mime_data(3), file using curl_mime_filedata(3) and user-defined data
 527read callback using curl_mime_data_cb(3). curl_mime_name(3) sets a part's
 528(i.e.: form field) name, while curl_mime_filename(3) fills in the remote
 529filename. With curl_mime_type(3), you can tell the MIME type of a part,
 530curl_mime_headers(3) allows defining the part's headers. When a multi-part
 531body is no longer needed, you can destroy it using curl_mime_free(3).
 532
 533The following example sets two simple text parts with plain textual contents,
 534and then a file with binary contents and uploads the whole thing.
 535
 536~~~c
 537    curl_mime *multipart = curl_mime_init(handle);
 538    curl_mimepart *part = curl_mime_addpart(multipart);
 539    curl_mime_name(part, "name");
 540    curl_mime_data(part, "daniel", CURL_ZERO_TERMINATED);
 541    part = curl_mime_addpart(multipart);
 542    curl_mime_name(part, "project");
 543    curl_mime_data(part, "curl", CURL_ZERO_TERMINATED);
 544    part = curl_mime_addpart(multipart);
 545    curl_mime_name(part, "logotype-image");
 546    curl_mime_filedata(part, "curl.png");
 547
 548    /* Set the form info */
 549    curl_easy_setopt(handle, CURLOPT_MIMEPOST, multipart);
 550
 551    curl_easy_perform(handle); /* post away! */
 552
 553    /* free the post data again */
 554    curl_mime_free(multipart);
 555~~~
 556
 557To post multiple files for a single form field, you must supply each file in
 558a separate part, all with the same field name. Although function
 559curl_mime_subparts(3) implements nested multi-parts, this way of
 560multiple files posting is deprecated by RFC 7578, chapter 4.3.
 561
 562To set the data source from an already opened FILE pointer, use:
 563
 564~~~c
 565    curl_mime_data_cb(part, filesize, (curl_read_callback)fread,
 566                      (curl_seek_callback)fseek, NULL, filepointer);
 567~~~
 568
 569A deprecated curl_formadd(3) function is still supported in libcurl.
 570It should however not be used anymore for new designs and programs using it
 571ought to be converted to the MIME API. It is however described here as an
 572aid to conversion.
 573
 574Using *curl_formadd*, you add parts to the form. When you are done adding
 575parts, you post the whole form.
 576
 577The MIME API example above is expressed as follows using this function:
 578
 579~~~c
 580    struct curl_httppost *post = NULL;
 581    struct curl_httppost *last = NULL;
 582    curl_formadd(&post, &last,
 583                 CURLFORM_COPYNAME, "name",
 584                 CURLFORM_COPYCONTENTS, "daniel", CURLFORM_END);
 585    curl_formadd(&post, &last,
 586                 CURLFORM_COPYNAME, "project",
 587                 CURLFORM_COPYCONTENTS, "curl", CURLFORM_END);
 588    curl_formadd(&post, &last,
 589                 CURLFORM_COPYNAME, "logotype-image",
 590                 CURLFORM_FILECONTENT, "curl.png", CURLFORM_END);
 591
 592    /* Set the form info */
 593    curl_easy_setopt(handle, CURLOPT_HTTPPOST, post);
 594
 595    curl_easy_perform(handle); /* post away! */
 596
 597    /* free the post data again */
 598    curl_formfree(post);
 599~~~
 600
 601Multipart formposts are chains of parts using MIME-style separators and
 602headers. It means that each one of these separate parts get a few headers set
 603that describe the individual content-type, size etc. To enable your
 604application to handicraft this formpost even more, libcurl allows you to
 605supply your own set of custom headers to such an individual form part. You can
 606of course supply headers to as many parts as you like, but this little example
 607shows how you set headers to one specific part when you add that to the post
 608handle:
 609
 610~~~c
 611    struct curl_slist *headers = NULL;
 612    headers = curl_slist_append(headers, "Content-Type: text/xml");
 613
 614    curl_formadd(&post, &last,
 615                 CURLFORM_COPYNAME, "logotype-image",
 616                 CURLFORM_FILECONTENT, "curl.xml",
 617                 CURLFORM_CONTENTHEADER, headers,
 618                 CURLFORM_END);
 619
 620    curl_easy_perform(handle); /* post away! */
 621
 622    curl_formfree(post); /* free post */
 623    curl_slist_free_all(headers); /* free custom header list */
 624~~~
 625
 626Since all options on an easy handle are "sticky", they remain the same until
 627changed even if you do call curl_easy_perform(3), you may need to tell curl to
 628go back to a plain GET request if you intend to do one as your next request.
 629You force an easy handle to go back to GET by using the CURLOPT_HTTPGET(3)
 630option:
 631
 632~~~c
 633    curl_easy_setopt(handle, CURLOPT_HTTPGET, 1L);
 634~~~
 635
 636Setting CURLOPT_POSTFIELDS(3) to "" or NULL does *not* stop libcurl from doing
 637a POST. It makes it POST without any data to send!
 638
 639# Converting from deprecated form API to MIME API
 640
 641Four rules have to be respected in building the multi-part:
 642
 643- The easy handle must be created before building the multi-part.
 644
 645- The multi-part is always created by a call to curl_mime_init(handle).
 646
 647- Each part is created by a call to curl_mime_addpart(multipart).
 648
 649- When complete, the multi-part must be bound to the easy handle using
 650CURLOPT_MIMEPOST(3) instead of CURLOPT_HTTPPOST(3).
 651
 652Here are some example of *curl_formadd* calls to MIME API sequences:
 653
 654~~~c
 655    curl_formadd(&post, &last,
 656                 CURLFORM_COPYNAME, "id",
 657                 CURLFORM_COPYCONTENTS, "daniel", CURLFORM_END);
 658                 CURLFORM_CONTENTHEADER, headers,
 659                 CURLFORM_END);
 660~~~
 661becomes:
 662~~~c
 663    part = curl_mime_addpart(multipart);
 664    curl_mime_name(part, "id");
 665    curl_mime_data(part, "daniel", CURL_ZERO_TERMINATED);
 666    curl_mime_headers(part, headers, FALSE);
 667~~~
 668
 669Setting the last curl_mime_headers(3) argument to TRUE would have caused
 670the headers to be automatically released upon destroyed the multi-part, thus
 671saving a clean-up call to curl_slist_free_all(3).
 672
 673~~~c
 674    curl_formadd(&post, &last,
 675                 CURLFORM_PTRNAME, "logotype-image",
 676                 CURLFORM_FILECONTENT, "-",
 677                 CURLFORM_END);
 678~~~
 679becomes:
 680~~~c
 681    part = curl_mime_addpart(multipart);
 682    curl_mime_name(part, "logotype-image");
 683    curl_mime_data_cb(part, (curl_off_t)-1, fread, fseek, NULL, stdin);
 684~~~
 685
 686curl_mime_name(3) always copies the field name. The special filename "-" is
 687not supported by curl_mime_filename(3): to read an open file, use a callback
 688source using fread(). The transfer is be chunk-encoded since the data size is
 689unknown.
 690
 691~~~c
 692    curl_formadd(&post, &last,
 693                 CURLFORM_COPYNAME, "datafile[]",
 694                 CURLFORM_FILE, "file1",
 695                 CURLFORM_FILE, "file2",
 696                 CURLFORM_END);
 697~~~
 698becomes:
 699~~~c
 700    part = curl_mime_addpart(multipart);
 701    curl_mime_name(part, "datafile[]");
 702    curl_mime_filedata(part, "file1");
 703    part = curl_mime_addpart(multipart);
 704    curl_mime_name(part, "datafile[]");
 705    curl_mime_filedata(part, "file2");
 706~~~
 707
 708The deprecated multipart/mixed implementation of multiple files field is
 709translated to two distinct parts with the same name.
 710
 711~~~c
 712    curl_easy_setopt(handle, CURLOPT_READFUNCTION, myreadfunc);
 713    curl_formadd(&post, &last,
 714                 CURLFORM_COPYNAME, "stream",
 715                 CURLFORM_STREAM, arg,
 716                 CURLFORM_CONTENTLEN, (curl_off_t)datasize,
 717                 CURLFORM_FILENAME, "archive.zip",
 718                 CURLFORM_CONTENTTYPE, "application/zip",
 719                 CURLFORM_END);
 720~~~
 721becomes:
 722~~~c
 723    part = curl_mime_addpart(multipart);
 724    curl_mime_name(part, "stream");
 725    curl_mime_data_cb(part, (curl_off_t)datasize,
 726                      myreadfunc, NULL, NULL, arg);
 727    curl_mime_filename(part, "archive.zip");
 728    curl_mime_type(part, "application/zip");
 729~~~
 730
 731CURLOPT_READFUNCTION(3) callback is not used: it is replace by directly
 732setting the part source data from the callback read function.
 733
 734~~~c
 735    curl_formadd(&post, &last,
 736                 CURLFORM_COPYNAME, "memfile",
 737                 CURLFORM_BUFFER, "memfile.bin",
 738                 CURLFORM_BUFFERPTR, databuffer,
 739                 CURLFORM_BUFFERLENGTH, (long)sizeof(databuffer),
 740                 CURLFORM_END);
 741~~~
 742becomes:
 743~~~c
 744    part = curl_mime_addpart(multipart);
 745    curl_mime_name(part, "memfile");
 746    curl_mime_data(part, databuffer, (curl_off_t)sizeof(databuffer));
 747    curl_mime_filename(part, "memfile.bin");
 748~~~
 749
 750curl_mime_data(3) always copies the initial data: data buffer is thus
 751free for immediate reuse.
 752
 753~~~c
 754    curl_formadd(&post, &last,
 755                 CURLFORM_COPYNAME, "message",
 756                 CURLFORM_FILECONTENT, "msg.txt",
 757                 CURLFORM_END);
 758~~~
 759becomes:
 760~~~c
 761    part = curl_mime_addpart(multipart);
 762    curl_mime_name(part, "message");
 763    curl_mime_filedata(part, "msg.txt");
 764    curl_mime_filename(part, NULL);
 765~~~
 766
 767Use of curl_mime_filedata(3) sets the remote filename as a side effect: it is
 768therefore necessary to clear it for *CURLFORM_FILECONTENT* emulation.
 769
 770# Showing Progress
 771
 772For historical and traditional reasons, libcurl has a built-in progress meter
 773that can be switched on and then makes it present a progress meter in your
 774terminal.
 775
 776Switch on the progress meter by, oddly enough, setting
 777CURLOPT_NOPROGRESS(3) to zero. This option is set to 1 by default.
 778
 779For most applications, the built-in progress meter is useless and what instead
 780is interesting is the ability to specify a progress callback. The function
 781pointer you pass to libcurl is then called on irregular intervals with
 782information about the current transfer.
 783
 784Set the progress callback by using CURLOPT_PROGRESSFUNCTION(3). Pass a pointer
 785to a function that matches this prototype:
 786
 787~~~c
 788    int progress_callback(void *clientp,
 789                          double dltotal,
 790                          double dlnow,
 791                          double ultotal,
 792                          double ulnow);
 793~~~
 794
 795If any of the input arguments is unknown, a 0 is provided. The first argument,
 796the 'clientp' is the pointer you pass to libcurl with
 797CURLOPT_PROGRESSDATA(3). libcurl does not touch it.
 798
 799# libcurl with C++
 800
 801There is only one thing to keep in mind when using C++ instead of C when
 802interfacing libcurl:
 803
 804The callbacks CANNOT be non-static class member functions
 805
 806Example C++ code:
 807
 808~~~c
 809    class AClass {
 810      static size_t write_data(char *ptr, size_t size, size_t nmemb,
 811                               void *ourpointer)
 812      {
 813        /* do what you want with the data */
 814      }
 815    }
 816~~~
 817
 818# Proxies
 819
 820What "proxy" means according to Merriam-Webster: "a person authorized to act
 821for another" but also "the agency, function, or office of a deputy who acts as
 822a substitute for another".
 823
 824Proxies are exceedingly common these days. Companies often only offer Internet
 825access to employees through their proxies. Network clients or user-agents ask
 826the proxy for documents, the proxy does the actual request and then it returns
 827them.
 828
 829libcurl supports SOCKS and HTTP proxies. When a given URL is wanted, libcurl
 830asks the proxy for it instead of trying to connect to the actual remote host
 831identified in the URL.
 832
 833If you are using a SOCKS proxy, you may find that libcurl does not support all
 834operations through it.
 835
 836For HTTP proxies: the fact that the proxy is an HTTP proxy puts certain
 837restrictions on what can actually happen. A requested URL that might not be a
 838HTTP URL is passed to the HTTP proxy to deliver back to libcurl. This happens
 839transparently, and an application may not need to know. I say "may", because
 840at times it is important to understand that all operations over an HTTP proxy
 841use the HTTP protocol. For example, you cannot invoke your own custom FTP
 842commands or even proper FTP directory listings.
 843
 844## Proxy Options
 845
 846To tell libcurl to use a proxy at a given port number:
 847~~~c
 848    curl_easy_setopt(handle, CURLOPT_PROXY, "proxy-host.com:8080");
 849~~~
 850Some proxies require user authentication before allowing a request, and you
 851pass that information similar to this:
 852~~~c
 853    curl_easy_setopt(handle, CURLOPT_PROXYUSERPWD, "user:password");
 854~~~
 855If you want to, you can specify the hostname only in the
 856CURLOPT_PROXY(3) option, and set the port number separately with
 857CURLOPT_PROXYPORT(3).
 858
 859Tell libcurl what kind of proxy it is with CURLOPT_PROXYTYPE(3) (if not,
 860it defaults to assuming an HTTP proxy):
 861~~~c
 862    curl_easy_setopt(handle, CURLOPT_PROXYTYPE, CURLPROXY_SOCKS4);
 863~~~
 864
 865## Environment Variables
 866
 867libcurl automatically checks and uses a set of environment variables to know
 868what proxies to use for certain protocols. The names of the variables are
 869following an old tradition and are built up as "[protocol]_proxy" (note the
 870lower casing). Which makes the variable 'http_proxy' checked for a name of a
 871proxy to use when the input URL is HTTP. Following the same rule, the variable
 872named 'ftp_proxy' is checked for FTP URLs. Again, the proxies are always HTTP
 873proxies, the different names of the variables allow different HTTP
 874proxies to be used.
 875
 876The proxy environment variable contents should be in the format
 877"[protocol://][user:password@]machine[:port]". Where the protocol:// part
 878specifies which type of proxy it is, and the optional port number specifies on
 879which port the proxy operates. If not specified, the internal default port
 880number is used and that is most likely not the one you would like it to be.
 881
 882There are two special environment variables. 'all_proxy' is what sets proxy
 883for any URL in case the protocol specific variable was not set, and 'no_proxy'
 884defines a list of hosts that should not use a proxy even though a variable may
 885say so. If 'no_proxy' is a plain asterisk ("*") it matches all hosts.
 886
 887To explicitly disable libcurl's checking for and using the proxy environment
 888variables, set the proxy name to "" - an empty string - with
 889CURLOPT_PROXY(3).
 890
 891## SSL and Proxies
 892
 893SSL is for secure point-to-point connections. This involves strong encryption
 894and similar things, which effectively makes it impossible for a proxy to
 895operate as a "man in between" which the proxy's task is, as previously
 896discussed. Instead, the only way to have SSL work over an HTTP proxy is to ask
 897the proxy to tunnel everything through without being able to check or fiddle
 898with the traffic.
 899
 900Opening an SSL connection over an HTTP proxy is therefore a matter of asking the
 901proxy for a straight connection to the target host on a specified port. This
 902is made with the HTTP request CONNECT. ("please dear proxy, connect me to that
 903remote host").
 904
 905Because of the nature of this operation, where the proxy has no idea what kind
 906of data that is passed in and out through this tunnel, this breaks some of the
 907few advantages that come from using a proxy, such as caching. Many
 908organizations prevent this kind of tunneling to other destination port numbers
 909than 443 (which is the default HTTPS port number).
 910
 911## Tunneling Through Proxy
 912
 913As explained above, tunneling is required for SSL to work and often even
 914restricted to the operation intended for SSL; HTTPS.
 915
 916This is however not the only time proxy-tunneling might offer benefits to
 917you or your application.
 918
 919As tunneling opens a direct connection from your application to the remote
 920machine, it suddenly also re-introduces the ability to do non-HTTP
 921operations over an HTTP proxy. You can in fact use things such as FTP
 922upload or FTP custom commands this way.
 923
 924Again, this is often prevented by the administrators of proxies and is
 925rarely allowed.
 926
 927Tell libcurl to use proxy tunneling like this:
 928~~~c
 929    curl_easy_setopt(handle, CURLOPT_HTTPPROXYTUNNEL, 1L);
 930~~~
 931In fact, there might even be times when you want to do plain HTTP operations
 932using a tunnel like this, as it then enables you to operate on the remote
 933server instead of asking the proxy to do so. libcurl does not stand in the way
 934for such innovative actions either!
 935
 936## Proxy Auto-Config
 937
 938Netscape first came up with this. It is a webpage (usually using a .pac
 939extension) with a JavaScript that when executed by the browser with the
 940requested URL as input, returns information to the browser on how to connect
 941to the URL. The returned information might be "DIRECT" (which means no proxy
 942should be used), "PROXY host:port" (to tell the browser where the proxy for
 943this particular URL is) or "SOCKS host:port" (to direct the browser to a SOCKS
 944proxy).
 945
 946libcurl has no means to interpret or evaluate JavaScript and thus it does not
 947support this. If you get yourself in a position where you face this nasty
 948invention, the following advice have been mentioned and used in the past:
 949
 950- Depending on the JavaScript complexity, write up a script that translates it
 951to another language and execute that.
 952
 953- Read the JavaScript code and rewrite the same logic in another language.
 954
 955- Implement a JavaScript interpreter; people have successfully used the
 956Mozilla JavaScript engine in the past.
 957
 958- Ask your admins to stop this, for a static proxy setup or similar.
 959
 960# Persistence Is The Way to Happiness
 961
 962Re-cycling the same easy handle several times when doing multiple requests is
 963the way to go.
 964
 965After each single curl_easy_perform(3) operation, libcurl keeps the connection
 966alive and open. A subsequent request using the same easy handle to the same
 967host might be able to reuse the already open connection! This reduces network
 968impact a lot.
 969
 970Even if the connection is dropped, all connections involving SSL to the same
 971host again, benefit from libcurl's session ID cache that drastically reduces
 972re-connection time.
 973
 974FTP connections that are kept alive save a lot of time, as the command-
 975response round-trips are skipped, and also you do not risk getting blocked
 976without permission to login again like on many FTP servers only allowing N
 977persons to be logged in at the same time.
 978
 979libcurl caches DNS name resolving results, to make lookups of a previously
 980looked up name a lot faster.
 981
 982Other interesting details that improve performance for subsequent requests
 983may also be added in the future.
 984
 985Each easy handle attempts to keep the last few connections alive for a while
 986in case they are to be used again. You can set the size of this "cache" with
 987the CURLOPT_MAXCONNECTS(3) option. Default is 5. There is rarely any point in
 988changing this value, and if you think of changing this it is often a reason to
 989think again.
 990
 991To force your upcoming request to not use an already existing connection, you
 992can do that by setting CURLOPT_FRESH_CONNECT(3) to 1. In a similar
 993spirit, you can also forbid the upcoming request to be "lying" around and
 994possibly get reused after the request by setting
 995CURLOPT_FORBID_REUSE(3) to 1.
 996
 997# HTTP Headers Used by libcurl
 998
 999When you use libcurl to do HTTP requests, it passes along a series of headers
1000automatically. It might be good for you to know and understand these. You can
1001replace or remove them by using the CURLOPT_HTTPHEADER(3) option.
1002
1003## Host
1004
1005This header is required by HTTP 1.1 and even many 1.0 servers and should be
1006the name of the server we want to talk to. This includes the port number if
1007anything but default.
1008
1009## Accept
1010
1011"*/*"
1012
1013## Expect
1014
1015When doing POST requests, libcurl sets this header to "100-continue" to ask
1016the server for an "OK" message before it proceeds with sending the data part
1017of the post. If the posted data amount is deemed "small", libcurl does not use
1018this header.
1019
1020# Customizing Operations
1021
1022There is an ongoing development today where more and more protocols are built
1023upon HTTP for transport. This has obvious benefits as HTTP is a tested and
1024reliable protocol that is widely deployed and has excellent proxy-support.
1025
1026When you use one of these protocols, and even when doing other kinds of
1027programming you may need to change the traditional HTTP (or FTP or...)
1028manners. You may need to change words, headers or various data.
1029
1030libcurl is your friend here too.
1031
1032## CURLOPT_CUSTOMREQUEST
1033
1034If changing the actual HTTP request keyword is what you want, like when GET,
1035HEAD or POST is not good enough for you, CURLOPT_CUSTOMREQUEST(3) is there for
1036you. It is simple to use:
1037
1038~~~c
1039    curl_easy_setopt(handle, CURLOPT_CUSTOMREQUEST, "MYOWNREQUEST");
1040~~~
1041
1042When using the custom request, you change the request keyword of the actual
1043request you are performing. Thus, by default you make a GET request but you
1044can also make a POST operation (as described before) and then replace the POST
1045keyword if you want to. You are the boss.
1046
1047## Modify Headers
1048
1049HTTP-like protocols pass a series of headers to the server when doing the
1050request, and you are free to pass any amount of extra headers that you
1051think fit. Adding headers is this easy:
1052
1053~~~c
1054    struct curl_slist *headers = NULL; /* init to NULL is important */
1055
1056    headers = curl_slist_append(headers, "Hey-server-hey: how are you?");
1057    headers = curl_slist_append(headers, "X-silly-content: yes");
1058
1059    /* pass our list of custom made headers */
1060    curl_easy_setopt(handle, CURLOPT_HTTPHEADER, headers);
1061
1062    curl_easy_perform(handle); /* transfer http */
1063
1064    curl_slist_free_all(headers); /* free the header list */
1065~~~
1066
1067... and if you think some of the internally generated headers, such as Accept:
1068or Host: do not contain the data you want them to contain, you can replace
1069them by setting them too:
1070
1071~~~c
1072    headers = curl_slist_append(headers, "Accept: Agent-007");
1073    headers = curl_slist_append(headers, "Host: munged.host.line");
1074~~~
1075
1076## Delete Headers
1077
1078If you replace an existing header with one with no contents, you prevent the
1079header from being sent. For instance, if you want to completely prevent the
1080"Accept:" header from being sent, you can disable it with code similar to
1081this:
1082
1083~~~c
1084    headers = curl_slist_append(headers, "Accept:");
1085~~~
1086
1087Both replacing and canceling internal headers should be done with careful
1088consideration and you should be aware that you may violate the HTTP protocol
1089when doing so.
1090
1091## Enforcing chunked transfer-encoding
1092
1093By making sure a request uses the custom header "Transfer-Encoding: chunked"
1094when doing a non-GET HTTP operation, libcurl switches over to "chunked"
1095upload, even though the size of the data to upload might be known. By default,
1096libcurl usually switches over to chunked upload automatically if the upload
1097data size is unknown.
1098
1099## HTTP Version
1100
1101All HTTP requests includes the version number to tell the server which version
1102we support. libcurl speaks HTTP 1.1 by default. Some old servers do not like
1103getting 1.1-requests and when dealing with stubborn old things like that, you
1104can tell libcurl to use 1.0 instead by doing something like this:
1105
1106~~~c
1107    curl_easy_setopt(handle, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_1_0);
1108~~~
1109
1110## FTP Custom Commands
1111
1112Not all protocols are HTTP-like, and thus the above may not help you when
1113you want to make, for example, your FTP transfers to behave differently.
1114
1115Sending custom commands to an FTP server means that you need to send the
1116commands exactly as the FTP server expects them (RFC 959 is a good guide
1117here), and you can only use commands that work on the control-connection
1118alone. All kinds of commands that require data interchange and thus need a
1119data-connection must be left to libcurl's own judgment. Also be aware that
1120libcurl does its best to change directory to the target directory before doing
1121any transfer, so if you change directory (with CWD or similar) you might
1122confuse libcurl and then it might not attempt to transfer the file in the
1123correct remote directory.
1124
1125A little example that deletes a given file before an operation:
1126
1127~~~c
1128    headers = curl_slist_append(headers, "DELE file-to-remove");
1129
1130    /* pass the list of custom commands to the handle */
1131    curl_easy_setopt(handle, CURLOPT_QUOTE, headers);
1132
1133    curl_easy_perform(handle); /* transfer ftp data! */
1134
1135    curl_slist_free_all(headers); /* free the header list */
1136~~~
1137
1138If you would instead want this operation (or chain of operations) to happen
1139_after_ the data transfer took place the option to curl_easy_setopt(3)
1140would instead be called CURLOPT_POSTQUOTE(3) and used the exact same
1141way.
1142
1143The custom FTP commands are issued to the server in the same order they are
1144added to the list, and if a command gets an error code returned back from the
1145server, no more commands are issued and libcurl bails out with an error code
1146(CURLE_QUOTE_ERROR). Note that if you use CURLOPT_QUOTE(3) to send
1147commands before a transfer, no transfer actually takes place when a quote
1148command has failed.
1149
1150If you set the CURLOPT_HEADER(3) to 1, you tell libcurl to get
1151information about the target file and output "headers" about it. The headers
1152are in "HTTP-style", looking like they do in HTTP.
1153
1154The option to enable headers or to run custom FTP commands may be useful to
1155combine with CURLOPT_NOBODY(3). If this option is set, no actual file
1156content transfer is performed.
1157
1158## FTP Custom CURLOPT_CUSTOMREQUEST
1159
1160If you do want to list the contents of an FTP directory using your own defined
1161FTP command, CURLOPT_CUSTOMREQUEST(3) does that. "NLST" is the default one for
1162listing directories but you are free to pass in your idea of a good
1163alternative.
1164
1165# Cookies Without Chocolate Chips
1166
1167In the HTTP sense, a cookie is a name with an associated value. A server sends
1168the name and value to the client, and expects it to get sent back on every
1169subsequent request to the server that matches the particular conditions set.
1170The conditions include that the domain name and path match and that the cookie
1171has not become too old.
1172
1173In real-world cases, servers send new cookies to replace existing ones to
1174update them. Server use cookies to "track" users and to keep "sessions".
1175
1176Cookies are sent from server to clients with the header Set-Cookie: and
1177they are sent from clients to servers with the Cookie: header.
1178
1179To send whatever cookie you want to a server, you can use CURLOPT_COOKIE(3) to
1180set a cookie string like this:
1181
1182~~~c
1183    curl_easy_setopt(handle, CURLOPT_COOKIE, "name1=var1; name2=var2;");
1184~~~
1185
1186In many cases, that is not enough. You might want to dynamically save whatever
1187cookies the remote server passes to you, and make sure those cookies are then
1188used accordingly on later requests.
1189
1190One way to do this, is to save all headers you receive in a plain file and
1191when you make a request, you tell libcurl to read the previous headers to
1192figure out which cookies to use. Set the header file to read cookies from with
1193CURLOPT_COOKIEFILE(3).
1194
1195The CURLOPT_COOKIEFILE(3) option also automatically enables the cookie parser
1196in libcurl. Until the cookie parser is enabled, libcurl does not parse or
1197understand incoming cookies and they are instead ignored. When the parser is
1198enabled the cookies are understood and the cookies are kept in memory and used
1199properly in subsequent requests when the same handle is used. Many times this
1200is enough, and you may not have to save the cookies to disk at all. Note that
1201the file you specify to CURLOPT_COOKIEFILE(3) does not have to exist to enable
1202the parser, so a common way to enable the parser and not read any cookies is
1203to use the name of a file you know does not exist.
1204
1205If you would rather use existing cookies that you have previously received
1206with your Netscape or Mozilla browsers, you can make libcurl use that cookie
1207file as input. The CURLOPT_COOKIEFILE(3) is used for that too, as
1208libcurl automatically finds out what kind of file it is and acts accordingly.
1209
1210Perhaps the most advanced cookie operation libcurl offers, is saving the
1211entire internal cookie state back into a Netscape/Mozilla formatted cookie
1212file. We call that the cookie-jar. When you set a filename with
1213CURLOPT_COOKIEJAR(3), that filename is created and all received cookies get
1214stored in it when curl_easy_cleanup(3) is called. This enables cookies to get
1215passed on properly between multiple handles without any information getting
1216lost.
1217
1218# FTP Peculiarities We Need
1219
1220FTP transfers use a second TCP/IP connection for the data transfer. This is
1221usually a fact you can forget and ignore but at times this detail comes back
1222to haunt you. libcurl offers several different ways to customize how the
1223second connection is being made.
1224
1225libcurl can either connect to the server a second time or tell the server to
1226connect back to it. The first option is the default and it is also what works
1227best for all the people behind firewalls, NATs or IP-masquerading setups.
1228libcurl then tells the server to open up a new port and wait for a second
1229connection. This is by default attempted with EPSV first, and if that does not
1230work it tries PASV instead. (EPSV is an extension to the original FTP spec
1231and does not exist nor work on all FTP servers.)
1232
1233You can prevent libcurl from first trying the EPSV command by setting
1234CURLOPT_FTP_USE_EPSV(3) to zero.
1235
1236In some cases, you want to have the server connect back to you for the second
1237connection. This might be when the server is perhaps behind a firewall or
1238something and only allows connections on a single port. libcurl then informs
1239the remote server which IP address and port number to connect to. This is made
1240with the CURLOPT_FTPPORT(3) option. If you set it to "-", libcurl uses your
1241system's "default IP address". If you want to use a particular IP, you can set
1242the full IP address, a hostname to resolve to an IP address or even a local
1243network interface name that libcurl gets the IP address from.
1244
1245When doing the "PORT" approach, libcurl attempts to use the EPRT and the LPRT
1246before trying PORT, as they work with more protocols. You can disable this
1247behavior by setting CURLOPT_FTP_USE_EPRT(3) to zero.
1248
1249# MIME API revisited for SMTP and IMAP
1250
1251In addition to support HTTP multi-part form fields, the MIME API can be used
1252to build structured email messages and send them via SMTP or append such
1253messages to IMAP directories.
1254
1255A structured email message may contain several parts: some are displayed
1256inline by the MUA, some are attachments. Parts can also be structured as
1257multi-part, for example to include another email message or to offer several
1258text formats alternatives. This can be nested to any level.
1259
1260To build such a message, you prepare the nth-level multi-part and then include
1261it as a source to the parent multi-part using function
1262curl_mime_subparts(3). Once it has been
1263bound to its parent multi-part, a nth-level multi-part belongs to it and
1264should not be freed explicitly.
1265
1266Email messages data is not supposed to be non-ASCII and line length is
1267limited: fortunately, some transfer encodings are defined by the standards to
1268support the transmission of such incompatible data. Function
1269curl_mime_encoder(3) tells a part that its source data must be encoded
1270before being sent. It also generates the corresponding header for that part.
1271If the part data you want to send is already encoded in such a scheme, do not
1272use this function (this would over-encode it), but explicitly set the
1273corresponding part header.
1274
1275Upon sending such a message, libcurl prepends it with the header list
1276set with CURLOPT_HTTPHEADER(3), as zero level mime part headers.
1277
1278Here is an example building an email message with an inline plain/html text
1279alternative and a file attachment encoded in base64:
1280
1281~~~c
1282    curl_mime *message = curl_mime_init(handle);
1283
1284    /* The inline part is an alternative proposing the html and the text
1285       versions of the email. */
1286    curl_mime *alt = curl_mime_init(handle);
1287
1288    /* HTML message. */
1289    curl_mimepart *part = curl_mime_addpart(alt);
1290    curl_mime_data(part, "<html><body><p>This is HTML</p></body></html>",
1291                         CURL_ZERO_TERMINATED);
1292    curl_mime_type(part, "text/html");
1293
1294    /* Text message. */
1295    part = curl_mime_addpart(alt);
1296    curl_mime_data(part, "This is plain text message",
1297                         CURL_ZERO_TERMINATED);
1298
1299    /* Create the inline part. */
1300    part = curl_mime_addpart(message);
1301    curl_mime_subparts(part, alt);
1302    curl_mime_type(part, "multipart/alternative");
1303    struct curl_slist *headers = curl_slist_append(NULL,
1304                      "Content-Disposition: inline");
1305    curl_mime_headers(part, headers, TRUE);
1306
1307    /* Add the attachment. */
1308    part = curl_mime_addpart(message);
1309    curl_mime_filedata(part, "manual.pdf");
1310    curl_mime_encoder(part, "base64");
1311
1312    /* Build the mail headers. */
1313    headers = curl_slist_append(NULL, "From: me@example.com");
1314    headers = curl_slist_append(headers, "To: you@example.com");
1315
1316    /* Set these into the easy handle. */
1317    curl_easy_setopt(handle, CURLOPT_HTTPHEADER, headers);
1318    curl_easy_setopt(handle, CURLOPT_MIMEPOST, mime);
1319~~~
1320
1321It should be noted that appending a message to an IMAP directory requires
1322the message size to be known prior upload. It is therefore not possible to
1323include parts with unknown data size in this context.
1324
1325# Headers Equal Fun
1326
1327Some protocols provide "headers", meta-data separated from the normal
1328data. These headers are by default not included in the normal data stream, but
1329you can make them appear in the data stream by setting CURLOPT_HEADER(3)
1330to 1.
1331
1332What might be even more useful, is libcurl's ability to separate the headers
1333from the data and thus make the callbacks differ. You can for example set a
1334different pointer to pass to the ordinary write callback by setting
1335CURLOPT_HEADERDATA(3).
1336
1337Or, you can set an entirely separate function to receive the headers, by using
1338CURLOPT_HEADERFUNCTION(3).
1339
1340The headers are passed to the callback function one by one, and you can
1341depend on that fact. It makes it easier for you to add custom header parsers
1342etc.
1343
1344"Headers" for FTP transfers equal all the FTP server responses. They are not
1345actually true headers, but in this case we pretend they are! ;-)
1346
1347# Post Transfer Information
1348
1349See curl_easy_getinfo(3).
1350
1351# The multi Interface
1352
1353The easy interface as described in detail in this document is a synchronous
1354interface that transfers one file at a time and does not return until it is
1355done.
1356
1357The multi interface, on the other hand, allows your program to transfer
1358multiple files in both directions at the same time, without forcing you to use
1359multiple threads. The name might make it seem that the multi interface is for
1360multi-threaded programs, but the truth is almost the reverse. The multi
1361interface allows a single-threaded application to perform the same kinds of
1362multiple, simultaneous transfers that multi-threaded programs can perform. It
1363allows many of the benefits of multi-threaded transfers without the complexity
1364of managing and synchronizing many threads.
1365
1366To complicate matters somewhat more, there are even two versions of the multi
1367interface. The event based one, also called multi_socket and the "normal one"
1368designed for using with select(). See the libcurl-multi.3 man page for details
1369on the multi_socket event based API, this description here is for the select()
1370oriented one.
1371
1372To use this interface, you are better off if you first understand the basics
1373of how to use the easy interface. The multi interface is a way to make
1374multiple transfers at the same time by adding up multiple easy handles into
1375a "multi stack".
1376
1377You create the easy handles you want, one for each concurrent transfer, and
1378you set all the options like you learned above, and then you create a multi
1379handle with curl_multi_init(3) and add all those easy handles to that multi
1380handle with curl_multi_add_handle(3).
1381
1382When you have added the handles you have for the moment (you can still add new
1383ones at any time), you start the transfers by calling
1384curl_multi_perform(3).
1385
1386curl_multi_perform(3) is asynchronous. It only performs what can be done
1387now and then return control to your program. It is designed to never
1388block. You need to keep calling the function until all transfers are
1389completed.
1390
1391The best usage of this interface is when you do a select() on all possible
1392file descriptors or sockets to know when to call libcurl again. This also
1393makes it easy for you to wait and respond to actions on your own application's
1394sockets/handles. You figure out what to select() for by using
1395curl_multi_fdset(3), that fills in a set of *fd_set* variables for
1396you with the particular file descriptors libcurl uses for the moment.
1397
1398When you then call select(), it returns when one of the file handles signal
1399action and you then call curl_multi_perform(3) to allow libcurl to do
1400what it wants to do. Take note that libcurl does also feature some time-out
1401code so we advise you to never use long timeouts on select() before you call
1402curl_multi_perform(3) again. curl_multi_timeout(3) is provided to
1403help you get a suitable timeout period.
1404
1405Another precaution you should use: always call curl_multi_fdset(3)
1406immediately before the select() call since the current set of file descriptors
1407may change in any curl function invoke.
1408
1409If you want to stop the transfer of one of the easy handles in the stack, you
1410can use curl_multi_remove_handle(3) to remove individual easy
1411handles. Remember that easy handles should be curl_easy_cleanup(3)ed.
1412
1413When a transfer within the multi stack has finished, the counter of running
1414transfers (as filled in by curl_multi_perform(3)) decreases. When the
1415number reaches zero, all transfers are done.
1416
1417curl_multi_info_read(3) can be used to get information about completed
1418transfers. It then returns the CURLcode for each easy transfer, to allow you
1419to figure out success on each individual transfer.
1420
1421# SSL, Certificates and Other Tricks
1422
1423[ seeding, passwords, keys, certificates, ENGINE, ca certs ]
1424
1425# Sharing Data Between Easy Handles
1426
1427You can share some data between easy handles when the easy interface is used,
1428and some data is share automatically when you use the multi interface.
1429
1430When you add easy handles to a multi handle, these easy handles automatically
1431share a lot of the data that otherwise would be kept on a per-easy handle
1432basis when the easy interface is used.
1433
1434The DNS cache is shared between handles within a multi handle, making
1435subsequent name resolving faster, and the connection pool that is kept to
1436better allow persistent connections and connection reuse is also shared. If
1437you are using the easy interface, you can still share these between specific
1438easy handles by using the share interface, see libcurl-share(3).
1439
1440Some things are never shared automatically, not within multi handles, like for
1441example cookies so the only way to share that is with the share interface.
1442
1443# Footnotes
1444
1445## [1]
1446
1447libcurl 7.10.3 and later have the ability to switch over to chunked
1448Transfer-Encoding in cases where HTTP uploads are done with data of an unknown
1449size.
1450
1451## [2]
1452
1453This happens on Windows machines when libcurl is built and used as a DLL. You
1454can still do this on Windows if you link with a static library.
1455
1456## [3]
1457
1458The curl-config tool is generated at build-time (on Unix-like systems) and
1459should be installed with the 'make install' or similar instruction that
1460installs the library, header files, man pages etc.
1461
1462## [4]
1463
1464This behavior was different in versions before 7.17.0, where strings had to
1465remain valid past the end of the curl_easy_setopt(3) call.