c++ web scripting

S

Shailesh Humbad

Has anyone ever heard of a c++ web 'scripting' engine? I think it
would be fairly easy to make one. All that is needed is an
intermediary that compiles the code on demand and caches the executables.

For example, I put a file like hello.cpp:

#include <iostream>
int main(void) {
cout << "<h1>hello world</h1>";
}

in the web server directory, and when I point my browser to
http://localhost/hello.cpp, the c++ web 'scripting' engine compiles
the code on the fly and serves the output over http. It would work
just like CGI, except the compilation is automatic.
 
J

John Harrison

Shailesh Humbad said:
Has anyone ever heard of a c++ web 'scripting' engine? I think it would
be fairly easy to make one. All that is needed is an intermediary that
compiles the code on demand and caches the executables.

For example, I put a file like hello.cpp:

#include <iostream>
int main(void) {
cout << "<h1>hello world</h1>";
}

in the web server directory, and when I point my browser to
http://localhost/hello.cpp, the c++ web 'scripting' engine compiles the
code on the fly and serves the output over http. It would work just like
CGI, except the compilation is automatic.
Click to expand...

It would be easy to setup a web server to do that. Your 'scripting engine'
is just the compiler followed by running the program.

john
 
P

Phlip

Shailesh said:
Has anyone ever heard of a c++ web 'scripting' engine?
Click to expand...

cgicc. It works great.
I think it
would be fairly easy to make one. All that is needed is an
intermediary that compiles the code on demand and caches the executables.
Click to expand...

Wha? What's wrong with just configuring your CGI to call your .EXE or
binary? I done all this, last millenium.
 
O

one2001boy

Shailesh said:
Has anyone ever heard of a c++ web 'scripting' engine? I think it would
be fairly easy to make one. All that is needed is an intermediary that
compiles the code on demand and caches the executables.

For example, I put a file like hello.cpp:

#include <iostream>
int main(void) {
cout << "<h1>hello world</h1>";
}

in the web server directory, and when I point my browser to
http://localhost/hello.cpp, the c++ web 'scripting' engine compiles the
Click to expand...

for CGI, it won't work this way for your above code.
code on the fly and serves the output over http. It would work just
like CGI, except the compilation is automatic.
Click to expand...

C++ cgi code can already work with C/C++ interpreter Ch.
 
O

Owen Jacobson

cgicc. It works great.
Click to expand...

Seconded. cgicc works way better if you want it to output compliant HTML
than if you want "normal" (nonstandard, table-driven, or otherwise
silly) HTML, though. IMO this is a feature, though from the mailing list I
seem to be in the minority.
 
T

Tommi =?UTF-8?B?TcOka2l0YWxv?=

Shailesh said:
Has anyone ever heard of a c++ web 'scripting' engine? I think it
would be fairly easy to make one. All that is needed is an
intermediary that compiles the code on demand and caches the executables.

For example, I put a file like hello.cpp:

#include <iostream>
int main(void) {
cout << "<h1>hello world</h1>";
}

in the web server directory, and when I point my browser to
http://localhost/hello.cpp, the c++ web 'scripting' engine compiles
the code on the fly and serves the output over http. It would work
just like CGI, except the compilation is automatic.
Click to expand...

Hi,

I have written a webapplicationserver in c++, which supports c++. I defined
a template-language like PHP, JSP or Mason for c++, so you can embed
c++-code in HTML-pages. It much faster than cgi, because there is no need
to start a process for each request. Even dynamic pages with this system
are faster than static pages in apache.

Here is a example of a dynamic web-page:

<%args>
some_text // define query-parameter here
</%args>
<html>
<body>
<# this is a comment, which is precompiled away #>
<{
// do some c++-coding here:
for (unsigned i = 0; i < 10; ++i)
{
// output the value of the query-param "some_text" 10 times:
}>
<p> <$ i $> : <$ some_text $> </p>
<{
} // end of for-loop
}>
</body>
</html>
__END__

The tags <$ ... $> outputs a c++-expression. Everything here is written to a
std::eek:stream, so you can output every object which has a
"operator<<(std::eek:stream&)" defined.

There are many more features like embedding other c++-components or special
tags for initialization or support for i18n.

I would like to release it under the GPL, but it lacks documentation yet.
What I have is a short introduction as a word-document written in german.
If someone is interested in helping, let me know.

It would be also nice to hear comments.


Tommi Mäkitalo
 
S

Shailesh Humbad

Tommi said:
Shailesh Humbad wrote:




Hi,

I have written a webapplicationserver in c++, which supports c++. I defined
a template-language like PHP, JSP or Mason for c++, so you can embed
c++-code in HTML-pages. It much faster than cgi, because there is no need
to start a process for each request. Even dynamic pages with this system
are faster than static pages in apache.

Here is a example of a dynamic web-page:

<%args>
some_text // define query-parameter here
</%args>
<html>
<body>
<# this is a comment, which is precompiled away #>
<{
// do some c++-coding here:
for (unsigned i = 0; i < 10; ++i)
{
// output the value of the query-param "some_text" 10 times:
}>
<p> <$ i $> : <$ some_text $> </p>
<{
} // end of for-loop
}>
</body>
</html>
__END__

The tags <$ ... $> outputs a c++-expression. Everything here is written to a
std::eek:stream, so you can output every object which has a
"operator<<(std::eek:stream&)" defined.

There are many more features like embedding other c++-components or special
tags for initialization or support for i18n.

I would like to release it under the GPL, but it lacks documentation yet.
What I have is a short introduction as a word-document written in german.
If someone is interested in helping, let me know.

It would be also nice to hear comments.


Tommi Mäkitalo
Click to expand...

Hi Tommi,

This is exactly what I was looking for. I'm so surprised nothing like
this already exists. If you think about it, .Net's "new and improved"
approach toward web applications is to compile the code on request,
and keep the compiled object code in a cache. However, .Net does not
compile to native code, but to an intermediate language, which adds
overhead. We were so busy with ASP, PHP, and other scripting
languages, that we only now realized we can use C++ directly and just
have the server compile it on the fly.

I understand that CGI is slow because of the need to create new
processes. I'd be interested to understand your implementation and
try it out.

My initial comment is that you have three different kinds of
delimeters, <{, <#, and <$. ASP uses <% while PHP uses <?. I think
the easiest to type is the last one, <?, because it only involves the
left-shift key with the left hand, and the lower, easily-reachable "?"
key with the right hand. Also, I'm not sure there is a need for
multiple kinds of delimeters. In ASP, when you want to print out a
value, you write <% =some_var %>, and it has no equivalent shortcut
for comments. PHP has no equivalent for either comments or printing
values, although you can leave out the semicolon for single line
statements, e.g. <?php print "hello" ?>. Now by default in PHP, the
opening tag is required to be <?php. Maybe you can use something like
this for your engine, like <?tomi ?>, and use standard commenting
paradigms for comments and values, like <?tomi /* this is comments */
?> and <?tomi cout << "hello"; ?>. These are just suggestions.

Shailesh
 
S

Shailesh Humbad

Owen said:
Seconded. cgicc works way better if you want it to output compliant HTML
than if you want "normal" (nonstandard, table-driven, or otherwise
silly) HTML, though. IMO this is a feature, though from the mailing list I
seem to be in the minority.
Click to expand...
(I posted this a few days ago, but it went to alt.html by accident.)

I tried cgicc prior to posting. The difference is the application
development cycle. I assume there are three steps with cgicc: edit
the code, run the compiler, and then reload the browser. When I say
scripting engine, I mean that there are only two steps: edit the code,
and reload the browser.

I see ch is an interpreter for a superset of C (with classes from
C++), but not full C++. Not only that, but I don't want
interpretation, I want cached, natively-compiled executables.

Like another poster said, it would be fairly easy to implement such a
thing. I'm just surprised it hasn't yet. Why use .Net for your web
projects when you can use C++, and skip the extra layer of complexity?
 
P

Phlip

Shailesh said:
I tried cgicc prior to posting. The difference is the application
development cycle. I assume there are three steps with cgicc: edit
the code, run the compiler, and then reload the browser. When I say
scripting engine, I mean that there are only two steps: edit the code,
and reload the browser.

I see ch is an interpreter for a superset of C (with classes from
C++), but not full C++. Not only that, but I don't want
interpretation, I want cached, natively-compiled executables.

Like another poster said, it would be fairly easy to implement such a
thing. I'm just surprised it hasn't yet. Why use .Net for your web
projects when you can use C++, and skip the extra layer of complexity?
Click to expand...

Why should C++ recompile between two page hits?

You need to think about routing data strictly thru XML and XSLT in XHTML
output mode. Then, regardless of the language substrate, you have templates
in a widely supported template language. Not a system that lets you couple
HTML to C++ in elaborate ways.
 
S

Shailesh Humbad

Phlip said:
Shailesh Humbad wrote:




Why should C++ recompile between two page hits?

You need to think about routing data strictly thru XML and XSLT in XHTML
output mode. Then, regardless of the language substrate, you have templates
in a widely supported template language. Not a system that lets you couple
HTML to C++ in elaborate ways.
Click to expand...

First, I think it was pretty obvious from my post, "I want *cached*,
natively-compiled executables", that the C++ engine would not
recompile between two hits to the same, unmodified page.

Second, given that not all data is stored in XML, simple XSLT
transformation is not always possible. We need access to databases,
hence libraries written in programming languages like C++. But this
point about where to put transformation logic seems more like a
religious debate.
 
D

David Hilsee

Shailesh Humbad said:
Hi Tommi,

This is exactly what I was looking for. I'm so surprised nothing like
this already exists. If you think about it, .Net's "new and improved"
approach toward web applications is to compile the code on request,
and keep the compiled object code in a cache. However, .Net does not
compile to native code, but to an intermediate language, which adds
overhead. We were so busy with ASP, PHP, and other scripting
languages, that we only now realized we can use C++ directly and just
have the server compile it on the fly.
Click to expand...
<snip>

Doesn't .Net provide one or two good ways to use C++ code inside a .Net
application? If not P/Invoke, then at least using "unsafe"? Then, if you
go with .Net, you could choose between higher-level languages and C++
depending on your needs.
 
P

Phlip

Shailesh said:
First, I think it was pretty obvious from my post, "I want *cached*,
natively-compiled executables", that the C++ engine would not
recompile between two hits to the same, unmodified page.
Click to expand...

Okay, this is a non-sequitur. Here's how C++ works. You write a program, run
a compiler, produce an executable, put it on a server, and stroll away. The
executable never changes. It may frequently serve similar pages with
different content. There's never a need to recompile it (regardless what
page content it serves).

When you say "cached", this makes me think you think C++ must recompile each
time it is used, or each time the data changes, or something. If you
envision some slight performance boost from serving the same page twice
without re-rendering it, then you are confusing some other CGI optimization
system with C++ compilation. If you need such a performance boost, just save
your page an re-serve it.

Please tell a complete, round-trip story that you envision some hypothetical
C++ CGI layer doing.
Second, given that not all data is stored in XML, simple XSLT
transformation is not always possible. We need access to databases,
hence libraries written in programming languages like C++. But this
point about where to put transformation logic seems more like a
religious debate.
Click to expand...

No religion needed. If you pipe data, whatever its format, into XML, it will
become compatible with a very good template system, XSLT, which has a
built-in mode to emit XHTML.

Further, you can detect the user's browser type, and if it is compatible,
you can send it your XSLT, and it will render the pages for you. All you
need to do is send that browser your XML payload, and it does the rest.
That's not religion, it is interoperability. We could have got it with
tab-delimited files, or INI file format, or whatever. As an accident of
history, we got it with XML.
 
S

Shailesh Humbad

Phlip said:
Okay, this is a non-sequitur. Here's how C++ works. You write a program, run
a compiler, produce an executable, put it on a server, and stroll away. The
executable never changes. It may frequently serve similar pages with
different content. There's never a need to recompile it (regardless what
page content it serves).

When you say "cached", this makes me think you think C++ must recompile each
time it is used, or each time the data changes, or something. If you
envision some slight performance boost from serving the same page twice
without re-rendering it, then you are confusing some other CGI optimization
system with C++ compilation. If you need such a performance boost, just save
your page an re-serve it.

Please tell a complete, round-trip story that you envision some hypothetical
C++ CGI layer doing.
Click to expand...

Yeah, I can see where the confusion lies. When I say cached, I mean
the executables themselves are cached, not the HTML output of those
executables. Here is the algorithm:

(In Web Server)
* Web server receives request for a .cpp file.
* Web server invokes the CGI handler for .cpp files, ServeCPPFile.
(In ServeCPPFile)
If file does not exist Then
Delete any file status entry from the database
Return 404
End If
* Lookup FileLastModifiedTime in database, searching by file name.
If no file status exist in the database Then
Save the FileName and LastModified timestamp in the database
Compile the cpp code (save exe in cache folder)
Else
If actual FileLastModified != FileLastModified in database Then
Compile the cpp code (save exe in cache folder)
End If
End If
* Run the compiled code
(In Web Server)
* Output of ServeCPPFile is sent to the client.

It's easiest to think of the C++ CGI Layer as spawning/forking a new
process to handle each page request. In this case, the ServeCPPFile
process is launched, and then ServeCPPFile launches another process to
run the compiled CPP code. In practice, some other techniques might
be used to save the overhead of creating processes.

Process heirarchy:
Apache -> ServeCPPFile -> EXE File

Output piping:
Client <- Apache <- ServeCPPFile <- EXE File
No religion needed. If you pipe data, whatever its format, into XML, it will
become compatible with a very good template system, XSLT, which has a
built-in mode to emit XHTML.

Further, you can detect the user's browser type, and if it is compatible,
you can send it your XSLT, and it will render the pages for you. All you
need to do is send that browser your XML payload, and it does the rest.
That's not religion, it is interoperability. We could have got it with
tab-delimited files, or INI file format, or whatever. As an accident of
history, we got it with XML.
Click to expand...

I'm not familiar with XSLT's advantages, having never tried it, but
I'll take your word for it. One caveat is that one loses some
performance for the gain in interoperability, particularly for legacy
data in databases.
 
T

Tommi =?UTF-8?B?TcOka2l0YWxv?=

....
<snip>
Click to expand...
(sorry - I would like to answer Shailesh comment, but my news-server missed
it)

Let me clearify the concept of my server first.

The idea is to create c++-code out of html-files and compile these into
something executable. I want to create compiled code. Just like I do
c++-programming. Not compile when called like Jsp, but when a client
requests a page, the code is here to generate the answer at full speed. No
compiler, no interpreter or runtime, which eats runtime or memory.

I tried to create a apache-module, but it failed, because apache is linked
with a c-linker and can't load shared libraries in c++. Next I thought
about creating a apache-module, which passes the requests to another
daemon, which is written in c++. But how do I pass the requests? A pipe
would be ok. I need to stream the request to my process and the process
streams the answer. But if my process waits for requests on a pipe, it can
also listen on a socket. So I had my webapplicationserver.

Html-pages with ebedded c++-code is precompiled into c++-classes, which is
compiled into object-code and linked to a shared library. The Webserver
loads this shared library dynamically and instantiates this classes to
process the received requests.

The precompiler, I wrote can even compile pictures like jpeg or gifs into
c++-classes, so I can put a whole webapplication including all graphics
into a single shared library.

The language, I created borrows much from Mason (http://www.masonhq.com)
because this was my preferred template-engine. That's the reason I don't
use tags from Jsp or Php or Asp.

Shailish suggested to reduce the tags to a single one like Php, which uses
<? or Asp with <%. It is one possible solution too. But I need more tags. I
started with <%some_tag>...</%some_tag> like Mason, which conformes to XML
too. In practice it is very hard to type, so I shortened some often used
tags to 2 characters manly to reduce typing and improve readability.
C++-code is marked with <%cpp>..</%cpp> or shorthand <{...}>, which
resembles a code-block in c++. Expressions are marked with <$...$> which is
like a perl- or shell-variable. Comments are <%doc>...</%doc> or shorthand
<#...#> like perl- or shell-comment. Another tag is <& ... &>, which embeds
another c++-component into the page, like a subroutine-call ('&' marks a
function in perl or a reference in c++).

I have other tags to define queryparameters, attributes, initialization,
defining subcomponents and others. This long form is extendable, when I
need something more.

And best of all: I have a working implementation.

OK - It is time to give it a name. The server is called 'tntnet'. The files,
with ebedded c++ are .ecpp-files.

I will publish my source very soon.


Tommi
 
P

Phlip

Yeah, I can see where the confusion lies. When I say cached, I mean
the executables themselves are cached, not the HTML output of those
executables. Here is the algorithm:

(In Web Server)
* Web server receives request for a .cpp file.
* Web server invokes the CGI handler for .cpp files, ServeCPPFile.
(In ServeCPPFile)
If file does not exist Then
Delete any file status entry from the database
Return 404
End If
* Lookup FileLastModifiedTime in database, searching by file name.
If no file status exist in the database Then
Save the FileName and LastModified timestamp in the database
Compile the cpp code (save exe in cache folder)
Else
If actual FileLastModified != FileLastModified in database Then
Compile the cpp code (save exe in cache folder)
End If
End If
* Run the compiled code
(In Web Server)
* Output of ServeCPPFile is sent to the client.

It's easiest to think of the C++ CGI Layer as spawning/forking a new
process to handle each page request. In this case, the ServeCPPFile
process is launched, and then ServeCPPFile launches another process to
run the compiled CPP code. In practice, some other techniques might
be used to save the overhead of creating processes.

Process heirarchy:
Apache -> ServeCPPFile -> EXE File

Output piping:
Client <- Apache <- ServeCPPFile <- EXE File
Click to expand...

You are planning to add an incredible amount of complexity and risk, without
benefit. What could such a Rube Goldberg contraption do that a text template
system can't do?
I'm not familiar with XSLT's advantages, having never tried it, but
I'll take your word for it. One caveat is that one loses some
performance for the gain in interoperability, particularly for legacy
data in databases.
Click to expand...

Premature optimization is the root of all evil.
 
O

Old Wolf

Shailesh Humbad said:
Like another poster said, it would be fairly easy to implement such a
thing. I'm just surprised it hasn't yet. Why use .Net for your web
projects when you can use C++, and skip the extra layer of complexity?
Click to expand...

Because in scripting languages there's lots of things you can
do in 1 line that require pages and pages of code and debugging in C++.
For example sockets. Why can't you go: fstream foo("http://bar.com");
and send your request and read the response with the << and >>
operators. It would be good if there were a bunch of libraries that
would enable you to write web-pages in C++ (compiled, I mean) without
having to do any low-level stuff. Another example would be parsing a
HTTP header and extracting any GET or POST variables.
 
S

Shailesh Humbad

Old said:
Because in scripting languages there's lots of things you can
do in 1 line that require pages and pages of code and debugging in C++.
For example sockets. Why can't you go: fstream foo("http://bar.com");
and send your request and read the response with the << and >>
operators. It would be good if there were a bunch of libraries that
would enable you to write web-pages in C++ (compiled, I mean) without
having to do any low-level stuff. Another example would be parsing a
HTTP header and extracting any GET or POST variables.
Click to expand...

I know exactly what you're talking about, having written some socket
apps in raw C/C++. The thing is that if those high-level libraries
don't already exist, they can be written. A lot of people have
invested their effort in creating easy-to-use web scripting libraries
(many written in C++ themselves!), but maybe they could have just
created those libraries for an embedded C++ engine in the web server.
I don't understand why we need to keep inventing new languages--it's
just a magnificent waste of limited educational resources.
 
P

Phlip

Shailesh said:
Can you tell me what XSLT engine you use? I would like to learn more
about it.
Click to expand...

Within C++, because I'm lazy, I use MSXML4. It was just lying in my system
folder, and I didn't need to download or install it. I just grab it with
this:

#import <msxml4.dll>

Don't write that. It gives me the same XML kit as other languages would
provide (possibly including .NET). However, if you don't know <comdef.h> or
other MS-specific C++ extensions to support COM, don't try to start now.
There are other libraries available but I don't know their names.

All websites should use only XHTML, because it's easier to test. Mistakes
are easier to catch. This line switches XSL to output XHTML:

<xsl:eek:utput method="xml" media-type="text/html"
standalone="no" omit-xml-declaration="yes"
encoding="UTF-8"/>

Here's a snip of useless XML:

<nodes>
<node>Hello</node>
<node>World</node>
</nodes>

Here is XSL to transform that into a Web page containing the payload inside
two text areas:

<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform">

<xsl:param name="fileName"/>

<xsl:eek:utput method="xml" media-type="text/html"
standalone="no" omit-xml-declaration="yes"/>

<xsl:template match="*">
<ul>
<li><xsl:value-of select="name()"/>
<ul>
<xsl:for-each select="*">
<li>
<form name="{$fileName}!/nodes/node[1]">
<xsl:value-of select="name()"/><TEXTAREA
name="contents">
<xsl:value-of select="text()"/></TEXTAREA>
</form>
</li>
</xsl:for-each>
</ul>
</li></ul>
</xsl:template>

</xsl:stylesheet>

The value of this system is HTML tags like <form> appears inline with <xsl>
stuff. The system matches escape character conventions, so programmers don't
need to convert them so often.

The drawback is the code looks like a sea of <><> characters!
 
T

Tommi =?UTF-8?B?TcOka2l0YWxv?=

Old said:
Because in scripting languages there's lots of things you can
do in 1 line that require pages and pages of code and debugging in C++.
For example sockets. Why can't you go: fstream foo("http://bar.com");
and send your request and read the response with the << and >>
operators. It would be good if there were a bunch of libraries that
would enable you to write web-pages in C++ (compiled, I mean) without
having to do any low-level stuff. Another example would be parsing a
HTTP header and extracting any GET or POST variables.
Click to expand...
Hi,

my webapplicationserver (tntnet) is just what you miss. You can write
web-pages in C++ (compiled, I mean) without having to do any low-level
stuff. It parses HTTP-headers, extracts GET or POST variables including
mime-decoding for http-uploads and offers them as std::string-variables to
the application.

It is multithreaded and completely written in c++. There is a precompiler to
compile html-pages into c++-classes.

Just write a webapplication it html with some special tags to escape to c++.
To generate output you can just stream (std::eek:stream) it to the client.

I use Linux for development, but every other OS, which supports the normal
Unix-API should work too. And it is multithreaded for best performance.

You can find it at http://www.maekitalo.de/tntnet. I hope to add some
documentation soon.


Tommi
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Similar Threads

scripting C++ 0
C as a web scripting language 32
C++ toolkit 2003: debugging programs 3
paxCompiler for scripting C++ applications 1
Installing Python as Scripting Language in IIS 3
Newbie with basic c++ question 19
C as a scripting language 88
Two C++ snippets for brainstorming 6

Members online

No members online now.
Total: 23 (members: 0, guests: 23)
Robots: 287

Forum statistics

Threads
473,755
Messages
2,569,536
Members
45,013
Latest member
KatriceSwa

Latest Threads

Top