Take the 2-minute tour ×
Programmers Stack Exchange is a question and answer site for professional programmers interested in conceptual questions about software development. It's 100% free, no registration required.

I'd like to get into web development using C++ as the "scripting language" on the server-side. My server infrastructure is *nix based, so doing web development in C++ on Azure is not applicable and C++/CLI ASP.NET is also not applicable.

Separate from legacy CGI applications, can web development be done using C++ ?


locked by Thomas Owens Jan 3 at 18:00

This question exists because it has historical significance, but it is not considered a good, on-topic question for this site, so please do not use it as evidence that you can ask similar questions here. This question and its answers are frozen and cannot be changed. More info: help center.

closed as too broad by gnat, GlenH7, MichaelT, Thomas Owens Jan 3 at 18:00

There are either too many possible answers, or good answers would be too long for this format. Please add details to narrow the answer set or to isolate an issue that can be answered in a few paragraphs. If this question can be reworded to fit the rules in the help center, please edit the question.

Of course it's possible, the question is; is it practical? –  Ed S. Mar 1 '11 at 22:44
See this question on stackoverflow.com. –  kevin cline Mar 1 '11 at 22:50
You could use assembly as a server-side language if you were so inclined. –  Channel72 Mar 2 '11 at 1:00
Or even Brainf*ck if , are . are redirected to a socket. –  dan04 Mar 2 '11 at 1:07
This brings back horrible memories of the first web project I was involved in. CGI gateways to C code. I still shudder when I think about it! :-) –  Brian Knoblauch Mar 2 '11 at 17:42

9 Answers 9

up vote 56 down vote accepted


There are even several frameworks for developing them, including Wt, cppcms, CSP, and others. FastCGI's mainline implementation is in C, and directly supports several languages, including C++.

Any programming language that can parse strings can be used in CGI or a servlet. Any language that can implement bindings with C libraries can also be used to develop modules for ISAPI- or Apache-compatible servers.

It's not particularly easy in C++, and good templating engines are few and far between, but it can be done.

Of course, the question of whether this is a good idea is another matter entirely. :)

Do note: Major websites like Amazon.com, eBay, and Google do use C++ for parts of their infrastructure. Realize, however, that Google only uses C++ for speed-critical systems, and Amazon.com only relatively recently switched away from Lisp (which angered some of their senior staff :).

Facebook formerly compiled PHP to C++, but their HipHop compiler (written partly in C++) has since been retooled as a bytecode virtual machine.

+1 For citing various frameworks. You should add that it's common for (very) big web apps to be powered by c++ (and other languages) : amazon.com, google.com, now facebook.com via hiphop, etc. –  Klaim Mar 1 '11 at 23:15
@Klaim: It's common, but it is by no means the rule. Amazon's architecture was historically Lisp-based and only recently rewritten in C++. Google's architecture involves Java, Python, and others almost as often as C++, all for various reasons. Facebook only uses hiphop now because they found out PHP doesn't scale. :) –  greyfade Mar 2 '11 at 0:22
I agree, but I meant that they still are well-known examples of use of C++ - to answer directly the original question title. –  Klaim Mar 2 '11 at 8:00
@johannes Facebook's scaling problem stems from the fact that they have to maintain an order of magnitude more servers than are otherwise necessary, specifically because of the poor performance of an optimized PHP script. Linear scaling simply isn't good enough for such a large infrastructure. But remember that the "shared nothing" approach is not exclusive to PHP. C and C++ can do that, too. –  greyfade May 17 '13 at 19:33
@amar The thing is there is little return except in the 0.1% of apps that need that raw performance. You could serve in 1/3 of the time in most other languages with good web stack support. Banks, web advertisers, etc all serve massive scale without resorting to C++. Even Facebook. Twitter. StackOverflow. All do it in higher level languages. Its here to stay but it isn't going to become the majority again. Probably ever. –  Rig Dec 16 '13 at 15:26

I guess that several embedded systems (e.g. routers, printers, ...) have some C++ driven web server.

In particular, you could use some HTTP server library like libonion to add some web capabilities to some C or C++ program, or to develop a light server with some web interface.

Some people are coding their Web server or their HTTP interface in Ocaml using Ocsigen. Not every web thing is PHP. And with FastCGI you could some dynamic web processing in/to your application.


Yes, it can be used. The others have mentioned various approaches. Here is my own approach. The advantage is that it is totally portable and self-contained, all the picked libraries only depend on ANSI C. Setting it up only requires the Linux Kernel and a C compiler (And the obvious stuff like Busybox, bash, etc) (or Windows and a compiler), no extra libraries are needed, no fancy huge installations.

The result is a single program which is both a web server and a dynamic page generator (Substitutes both "apache" and "php"), it will also have database access via sqlite.

Libraries used:

  • Mongoose - Http server
  • Sqlite - SQL Database
  • MiniXML - Makes dynamic page generation easier. sort of like Javascript's createElement

The rest of this answer is a complete set-up guide for Linux. Both SQlite and MiniXML are optional, but the guide covers the full installation. It is up to you to comment out the non needed parts if you're interested in disabling either sqlite or MiniXML.

1. Download the 3 libraries

2. Prepare your folder

  • Create an empty folder (We'll call it the main folder)
  • Put the following files in it:
    • From the sqlite tar.gz: sqlite3.c , sqlite3.h
    • From the Mongoose zip: mongoose.c , mongoose.h
    • From the mxml tar.gz: mxml.h

3. Compile mxml

You may have noticed mxml.c is missing, this is because we need to create a static mxml library. Go to the folder where the mxml tar.gz was downloaded and perform:

tar -xvf mxml-<version>.tar.gz #Extract the tar
cd mxml-<version> #Go to the newly extracted directory
./configure #prepare the compiler
make #compile, you may need to install "make" first.

Once the compilation is finished, many files will be generated, the only file of interest to us is libmxml.a, copy that file into the main folder.

3.1 Doublecheck

Check that the main folder has the following:

  • For mongoose: mongoose.c, mongoose.h
  • For mxml: libmxml.a, mxml.h
  • for sqlite: sqlite.c, sqlite.h

4. main.c

Let's create the actual program, create a main.c file in the main folder, here is a skeleton for you to get started.

#include <string.h>
#include <stdio.h>

#include "mongoose.h"
#include "mxml.h"
#include "sqlite3.h"

/***Sqlite initialization stuff***/
//comment out everything sqlite related if you don't want sqlite, including the callback function and the include "sqlite3.h"
static int callback(void * custom, int argc, char **argv, char **azColName);
char *zErrMsg = 0;
sqlite3 *db;
int rc;

/***Just some laziness shortcut functions I made***/
typedef mxml_node_t * dom; //type "dom" instead of "mxml_node_t *"
#define c mxmlNewElement   //type "c" instead of "mxmlNewElement"
inline void t(dom parent,const char *string) //etc
    mxmlNewText(parent, 0, string);

//type "sa" instead of "mxmlElementSetAttr"
inline void sa(dom element,const char * attribute,const char * value) 

//The only non boilerplate code around in this program is this function
void serve_hello_page(struct mg_connection *conn)
    char output[1000];
    mg_send_header(conn,"Content-Type","text/html; charset=utf-8");
    mg_printf_data(conn, "%s", "<!DOCTYPE html>");
    //This literally prints into the html document

    /*Let's generate some html, we could have avoided the
     * xml parser and just spat out pure html with mg_printf_data
     * e.g. mg_printF_data(conn,"%s", "<html>hello</html>") */

    //...But xml is cleaner, here we go:
            dom html=mxmlNewElement(MXML_NO_PARENT,"html");
                dom head=c(html,"head");
                    dom meta=c(head,"meta");
                dom body=c(html,"body");
                    t(body,"Hello, world<<"); //The < is auto escaped, neat!
                    t(body,"Fred ate bred");    
                dom table=c(body,"table");

                //populate the table via sqlite
                rc = sqlite3_exec(db, "SELECT * from myCoolTable", callback, table, &zErrMsg);
                if( rc!=SQLITE_OK )
                    fprintf(stderr, "SQL error: %s\n", zErrMsg);

            mxmlSaveString (html,output,1000,  MXML_NO_CALLBACK);
            mg_printf_data(conn, "%s", output);

//sqlite callback
static int callback(void * custom, int argc, char **argv, char **azColName)
    //this function is executed for each row
    dom table=(dom)custom;

    dom tr=c(table,"tr");
    dom td;
    int i;
    for(i=0; i<argc; i++)
        if (argv[i])
            t(td, argv[i]);
            t(td, "NULL");

        printf("%s == %s\n", azColName[i], argv[i] ? argv[i] : "NULL");
     return 0;

static int event_handler(struct mg_connection *conn, enum mg_event ev)
    if (ev == MG_AUTH)
        return MG_TRUE;   // Authorize all requests
    else if (ev == MG_REQUEST)
        if (!strcmp(conn->uri, "/hello"))
            return MG_TRUE;   // Mark as processed
    return MG_FALSE;  // Rest of the events are not processed


int main(void)
    struct mg_server *server = mg_create_server(NULL, event_handler);
    //mg_set_option(server, "document_root", "."); //prevent dir listing and auto file serving
    //TODO can I allow file listing without dir listing in a specified directory?
    mg_set_option(server, "listening_port", "8080");

    rc = sqlite3_open("db.sqlite3", &db); 

    if( rc )
        fprintf(stderr, "Can't open database: %s\n", sqlite3_errmsg(db));

    printf("Server is running on port 8080!\n");
    for (;;)
        mg_poll_server(server, 1000);  // Infinite loop, Ctrl-C to stop

    return 0;

 * useful stuff:
 * mg_send_file(struct mg_connection *, const char *path); - serve the file at *path*/

Finally, compiling!

Let's compile. cd to your main folder and execute these:

gcc -c main.c
gcc -c mongoose.c
gcc -c sqlite3.c
gcc -o server.out main.o mongoose.o sqlite3.o -ldl -lpthread -lmxml -L . 

Now, execute server.out with /server.out, and navigate to localhost:8080/hello

Done :)

github.com/bel2125/civetweb –  Hey Jan 3 at 16:13
@Hey: Thanks for pointing out this Mongoose alternative, I always prefer community-driven projects. I will probably replace Mongoose with Civetweb in my answer after I thoroughly test it. –  Hello World Jan 3 at 17:56

If you're planning to write your web application in C++, it would be total waste to then interface it as CGI.

My suggestion would be to build it asynchronous using ASIO (Asynchronous I/O). With that you can build blazing fast web service (combine with nginx as a reverse-proxy and statics server for best effects); Combine that with template library like Wt and you're ready to serve tens of thousands request per second from a single server.

Whether this is practical alternative to dynamic language web framework is another issue.


For PHP you can write your own C/C++ extensions and get good performance benefits that way. If I had a really CPU intensive part of my web application I would probably make a small C++ library that offloaded that processing to the extension and then returned the result back to the PHP and then the PHP outputs it to the browser.

The other thing people don't often consider is offloading certain CPU processing to the client side e.g. JavaScript/jQuery. If I've got a web server, I might need a a 3Ghz CPU to do CPU intensive processing for a particular function (maybe some data processing). My company is paying money for that server each month to keep it running. If I want to scale up operations for a 100 concurrent users running that CPU intensive task at the same time then maybe I need multiple CPUs and servers, increasing the cost to my business. If I offload that CPU intensive task to the client side, then each user that visits the website can do their own processing on the data and I don't have to increase my server capability therefore saving me money.

After all with all with the collective power of 100+ desktops/tablets/mobiles doing the processing for you that's a lot more power than your server sitting in a datacenter somewhere costing your business money each month to keep running. Potentially then all your server would be doing would be retrieving data from the database, serving content and a bit of pre/post processing and validation of the data before storing back in the database. Obviously you wouldn't make the client side code too CPU intensive which might block/freeze the web browser UI, you might fire off an AJAX request to the server, retrieve the data and then process the data asynchronously client-side, leaving the web-browser UI completely usable.


It appears Microsoft thinks it can too. Check out Casablanca which is a new set of tooling for (it appears) Azure using C++.

Casablanca is a project to start exploring how to best support C++ developers who want to take advantage of the radical shift in software architecture that cloud computing represents.

Here’s what you get with Casablanca:

  • Support for accessing REST services from native code on Windows Vista, Windows 7, and Windows 8 Consumer Preview by providing asynchronous C++ bindings to HTTP, JSON, and URIs
  • A Visual Studio extension SDK to help you write C++ HTTP client side code in your Windows 8 Metro style app
  • Support for writing native-code REST for Azure, including Visual Studio integration
  • Convenient libraries for accessing Azure blob and queue storage from native clients as a first class Platform-as-a-Service (PaaS) feature
  • A consistent and powerful model for composing asynchronous operations based on C++ 11 features
  • A C++ implementation of the Erlang actor-based programming model
  • A set of samples and documentation

in the beginning, it was quite common - the first web sites I worked on in the late 1990s were ISAPI extensions written in C++, and they worked quite well.

isapi.dll anyone? –  red-dirt Mar 5 '11 at 23:10
or ATLServer - atlserver.codeplex.com –  gbjbaanb Apr 27 '12 at 23:41

The short answer is, ANYTHING can be used to write a webpage provide it can read the input, write interpretable output, and is executable by the webserver.

Technically, any language can be used as a CGI script provided it:

  1. Interprets all the inputs and environment as presented by the server
  2. Outputs in a known markup language (generally html)
  3. Can be ran by the server

There are also other ways too. Perl has the ability to be built as a wrapper around c/c++ code, acting as an interpreting layer between the two (and this is not including perl modules that are flat out compiled as C).


Why not?

The OkCupid dating site is created with C++. There are probably other examples.

There's also a Qt-inspired toolkit for developing web applications with C++ called Wt.

"Why not"? Because it's much easier using a language that has more support for this sort of thing. –  Ed S. Mar 1 '11 at 22:47
@Ed S. As I and greyfade pointed out there are frameworks for developing web applications with C++. –  Vitor Mar 1 '11 at 22:49
Yes, but again, are they as easy to use as more commonly used frameworks? I'm honestly asking, I'm not a web developer and I've never used them, but something tells me they are likely not as mature or widely used as (for example) their ruby/python/PHP counterparts. –  Ed S. Mar 1 '11 at 22:55
@EdS.: Neither Ruby nor Python started with web frameworks. In fact it took a decade for those to appear. The frameworks are the mere consequence of enough people wanting to use language X for problem Y. The same could happen for C++. Main reasons why it didn't: C++ is not managed, takes ages to compile and has a higher entrance barrier in general. –  back2dos Nov 21 '13 at 20:20
@back2dos: Who said either language was developed with the web in mind? I certainly did not. I used the term "support". –  Ed S. Nov 21 '13 at 23:33

Not the answer you're looking for? Browse other questions tagged or ask your own question.