Run Perl in the browser with WebPerl!

View the Project on GitHub haukex/webperl

[ Using - Building - 🦋 - Notes - Legal - Wiki ]

Using WebPerl

Notice: WebPerl is very much in beta. Some things may not work yet, and parts of the API may still change. Your feedback is always appreciated!

This page documents the Perl 5 support, for the experimental Perl 6 support, see here.

Basic Usage

Getting WebPerl

If you plan on building WebPerl, for example if you’d like to add more CPAN modules, then head on over to Building WebPerl. Otherwise, if you’d just like to get started quickly and work with the prebuilt WebPerl (includes many of the Perl core modules plus a couple extras), then download webperl_prebuilt_v0.09-beta.zip and unpack it. This ZIP file includes the contents of the web directory of the source code, as well as the build products emperl.* (currently three files). If you’d like to work with the source code as checked out from GitHub, then you can copy these emperl.* files into the web directory of the source tree.

Serving WebPerl

You should serve WebPerl via a webserver of your choice, or you can use the included simple webperl.psgi for testing. You can run it using plackup from Plack by simply saying plackup webperl.psgi.

The following four files make up WebPerl:

I strongly recommend you add a MIME type of application/wasm for .wasm files, otherwise you may see warnings like “wasm streaming compile failed: TypeError: Response has unsupported MIME type” and “falling back to ArrayBuffer instantiation”. For example, in an Apache .htaccess file, you can say: AddType application/wasm .wasm

Note that opening the files locally (via file://) may not work due to browsers’ Same-Origin Policy. However, there are some workarounds:

See also the Emscripten deployment notes at http://kripken.github.io/emscripten-site/docs/compiling/Deploying-Pages.html, in particular I’d recommended using gzip encoding to serve the WebPerl files.

Including Perl code in your HTML

In your HTML file, add the following (usually inside the <head> tags):

<script src="webperl.js"></script>

Then, you can add one or more <script type="text/perl"> tags containing embedded Perl code, or a single <script type="text/perl" src="foo.pl"></script> tag which loads a Perl script from the server - but not both! The code from multiple <script type="text/perl"> tags will be concatenated and run as a single script.

If you use embedded <script type="text/perl"> tags, then the function js from WebPerl.pm will be imported automatically. If you want to customize the import list, then add use WebPerl ...; as one of the first five lines of your Perl code (to be exact, WebPerl will look for /^\s*use\s+WebPerl(\s|;)/m).

If you don’t have any such script tags in the document, Perl won’t be run automatically, and you can control Perl in detail via the JavaScript Perl object provided by webperl.js.

Note that unlike JavaScript, which is run immediately, WebPerl will always be loaded and run asynchronously from the page load. If you use <script type="text/perl"> tags, these will always be run after the document is ready, and if you use the Perl object as described below, you will have control over when Perl is initialized and run, but it will still be asynchronous because files need to be fetched from the server.

The Perl Interpreter and its Environment

The perl compiled for WebPerl is mostly a standard build of Perl, except for a few patches to make things compile properly, and the major differences described here.

Emscripten provides emulation for a number of system calls, most notably for WebPerl, it provides a virtual filesystem (details below) from which Perl can load its modules, since of course JavaScript in the browser is a sandboxed environment (no access to hardware, the local filesystem, etc.). However, because Perl is the only Emscripten process running in the browser, there are several things that won’t work either because Emscripten doesn’t support them (yet) or because they are simply not possible in this single-process environment:

Like many UI frameworks, scripting in the browser is usually asynchronous and event-driven. In addition, in Emscripten it is currently not easy to run a program multiple times. In order to better support these circumstances, WebPerl’s C main() function has been patched to not end the runtime. This means that once the main Perl script is run, the interpreter is not shut down, meaning END blocks and global destruction are not run, and instead control is passed back to the browser.

This way, you can write Perl in an event-driven manner: in your main code, you can register callbacks as event handlers for events such as button clicks, network communication, etc., and then control is passed back to the browser. When the events occur, your Perl callbacks will be run.

In order to allow for this mode of execution, WebPerl is built with Emscripten’s NO_EXIT_RUNTIME option enabled. When this option is enabled, atexit handlers are not supported, and calls to exit will result in a warning. For this reason, WebPerl is patched to not call exit when the exit code is zero. As a result of all this, in your scripts, I strongly recommend you don’t use Perl’s exit;/exit(0);, as it will not likely do what you want.

Remember that in the browser, the user may leave a page at any time, and there is little a script can do to prevent this. Although it’s possible to ask Perl to end early as follows, I would still recommend that you don’t rely on END blocks or global destruction. If your program is doing things like saving files (e.g. via an asynchronous network request), then you should provide some kind of feedback to your user to know that a process is still going on, and possibly install your own “beforeunload” handler.

WebPerl includes a C function int emperl_end_perl() which will perform the normal Perl interpreter shutdown (but as mentioned above, not call exit if the exit code is zero). This function is accessible in several ways:

These options might be useful if you’re porting an existing script to run in WebPerl.

(In addition, WebPerl currently registers an “beforeunload” handler that attempts to call the “end” function, but since this will be happening as the page is being unloaded, do not rely on this being able to do very much, or even being called at all!)

Memory Management and Anonymous subs

Anonymous subs passed from Perl to JavaScript must be explicitly freed when you are done using them, or else this is a memory leak. Please read this section!

When JavaScript arrays, objects, and functions are passed to Perl, they are not copied, instead they are given an ID and placed in a table so that when Perl wants to access them, it only needs to remember the ID, and pass the ID and the corresponding operation to JavaScript. In JavaScript, these objects are kept alive because of the entry in the table. Once the object goes out of scope in Perl, its DESTROY method lets JavaScript know that it can free that entry from the table, so JavaScript is free to garbage collect it if there are no other references.

When Perl values are passed to JavaScript, they are generally copied, except for anonymous subs, where a mechanism similar to the above is used, and a reference to the subs is kept alive using a table in Perl. However, JavaScript has no equivalent of the DESTROY method, which means that even if you are done using a sub in JavaScript, Perl will not know when it can free the table entry, unless you explicitly tell it to!

WebPerl provides two mechanisms for freeing subs:

Of course, it is often the case that anonymous subs need to persist for the entire run of a program (like for example click handlers for buttons), or that you may only have a handful of anonymous subs in your program overall. In such cases, you probably don’t need to unregister them. However, there are cases where this is very important to keep in mind - for example anonymous subs generated via stringy evals.

If you want to check how many anonymous subs are registered, you can say print scalar(keys %WebPerl::CodeTable); (do not modify this hash).

Note that the above only applies to anonymous subs. subs that exist in Perl’s symbol table will persist in Perl’s memory anyway, and no table entry is generated for them, because it is assumed you won’t delete them from the symbol table - so please don’t do that. Also, don’t rename or redefine subs after having passed them to JavaScript, as that will probably cause mysterious behavior.

Virtual Filesystem

Emscripten provides a virtual file system that also provides a few “fake” files such as /home/web_user, /dev, and others, so that it resembles a normal *NIX file system. This filesystem resides entirely in memory in the browser.

Perl’s libraries (*.pm) are installed into this virtual file system at /opt/perl. Note that because the perl binary is compiled to WebAssembly and XS libraries are statically linked into it, you won’t find any perl binary or shared library files in the virtual file system, or for that matter any other binaries, since this is a single-process environment.

It is important to keep apart the different ways to access files:

While a WebPerl instance is running, you can modify files in the virtual file system as you might be used to from regular Perl. But the virtual filesystem is reloaded every time WebPerl is reloaded, so any changes are lost! The exception is the “IDBFS” provided by Emscripten, which stores files in an IndexedDB, so they typically persist in the browser’s storage across sessions. WebPerl mounts an instance of this filesystem at /mnt/idb, which you are free to use. If you want your Perl script to write to files there, you must also use Emscripten’s FS.syncfs() interface after writing files, for example:

js(q/ FS.syncfs(false, function (err) {
	if(err) alert("FS sync failed: "+err);
	else console.log("FS sync ok"); }); /);

Remember that users may clear this storage at any time, so it is not really a permanent storage either. If you need to safely store files, it’s best to store them on the user’s machine (or the web server, if they are different machines) using one of the methods described above.

In particular, even though you might make heavy use of /mnt/idb when testing with the “mini IDE”, remember that this storage is not a way to distribute files to your users, and in fact, some users’ browsers may automatically regularly clear the IndexedDB, or have it disabled altogether. It also may not work at all in a “sandboxed” iframe. For providing your application to your users, either use <script type="text/perl"> tags, compile the script into the virtual file system, or use the JavaScript Perl object.

You might also put files into the virtual filesystem more permanently by modifying the “make install” step of build.pl. Keep in mind that like the other files in the virtual file system, any modifications will be lost once WebPerl is reloaded, and the only way to modify them is re-running build.pl.

Additional information on the virtual file system may be found at:

Note that WebPerl’s build process strips any POD from the Perl libraries, to reduce download size.

By the way, I don’t recommend relying on the initial working directory when WebPerl starts; either chdir to a known location, or always use absolute filenames.


webperl.js provides a JavaScript object Perl that can be used to control the Perl interpreter. Many properties of this object are intended for internal use by WebPerl only, so please only use the interface documented here.

Controlling Perl

As documented above, if your HTML file contains <script type="text/perl"> tags, these will be run automatically, so you should not use Perl.init() and Perl.start() in this case.


Initializes the Perl interpreter (asynchronously fetches the emperl.* files). You should pass this function a callback function, which is to be called when Perl is ready to be run - normally you would call Perl.start() from this callback.


Runs Perl with the given argv array. If argv is not provided, uses Emscripten’s Module.arguments, which currently defaults to ['--version'].


Evaluates the given Perl code. Currently always returns a string.

The functionality of this function may be expanded upon in the future to return more than just a string. See the discussion in Mappings from Perl to JavaScript.


Ends the Perl interpreter. See the discussion under “The Perl Interpreter and its Environment” for details.



Set this to a function (str,chan) {...} to handle Perl writing to STDOUT or STDERR. str is the string to be written, which may consist of a single character, a whole line, or multiple lines. chan will be either 1 for STDOUT or 2 for STDERR. If you want to merge the two streams, you can simply ignore the chan argument. Defaults to an implementation that line-buffers and logs via console.log(), prefixing either STDOUT or STDERR depending on the channel. See also Perl.makeOutputTextarea, which installs a different output handler.


If set to true before calling Perl.init(), then WebPerl will automatically end the Perl interpreter after it finishes running the main script. See the discussion under “The Perl Interpreter and its Environment”. Defaults to false.


If set to true before calling Perl.start(), then WebPerl will not automatically mount the IDBFS filesystem (see “Virtual File System”. Defaults to false.

This option was added in v0.05-beta.


Enable this option at any time to get additional trace-level output to console.debug(). Defaults to false.


Pass this function a function (from,to) {...} to register a new handler for state changes of the Perl interpreter.

The states currently are:

This function was added in WebPerl v0.05-beta.


This property should be read only! After Perl’s state has changed to Ended, you can retrieve the exit code here.

This property was added in WebPerl v0.09-beta.


Deprecated in WebPerl v0.05-beta. Use Perl.addStateChangeListener instead.

Set this to a function (from,to) {...} to handle state changes of the Perl interpreter. Defaults to a simple implementation that logs via console.debug().

Utility Functions


This function will create a new DOM <textarea> element, set up a Perl.output handler that redirects Perl’s output (merged STDOUT and STDERR) into the <textarea>, and return the DOM element. You may optionally pass this function a string argument giving a DOM ID. You will need to add the <textarea> to your DOM yourself (see webperl_demo.html for an example).


WebPerl.pm provides the Perl side of the WebPerl API. Its central function is js(), documented below. It also provides the functions unregister, sub_once, and sub1 (the latter two are aliases for each other), which are documented in “Memory Management and Anonymous subs”. For convenience, it can also re-export encode_json, so you can request it directly from WebPerl instead of needing to use another module. Additional functions, like js_new(), are documented below. All functions are exported only on request.

Note that WebPerl will also enable autoflush for STDOUT.


This function takes a single string argument consisting of JavaScript code to run, uses JavaScript’s eval to run it, and returns the result, as follows.

You may also pass an arrayref, hashref, or coderef, and this data structure will be passed to JavaScript, and a corresponding WebPerl::JSObject returned. Other references, including objects, are currently not supported.

Mappings from JavaScript to Perl

If the code given to js() throws a JavaScript error, js() will die. Otherwise, the js() function will return:


A WebPerl::JSObject is a thin wrapper around a JavaScript object. The contents of the JavaScript object are not copied to Perl, they are kept in JavaScript and accessed only when requested from Perl.

JSObjects support overload array, hash, and code dereferencing, plus autoloaded method calls. This means that if you have a WebPerl::JSObject stored in a Perl scalar $foo pointing to a JavaScript object foo:

JSObjects provide the following methods:

Method autoloading will of course not work for JavaScript methods that have the same name as existing Perl methods - these are the above methods, plus methods named AUTOLOAD, DESTROY, plus any methods inherited from Perl’s UNIVERSAL class, such as can or isa. If you need to call JavaScript methods with any of these names, use methodcall. For example, $jsobject->methodcall("can", "arg1") will call the JavaScript method can instead of the Perl method can.

Arguments from Perl to JavaScript function or method calls are mapped as follows.

Mappings from Perl to JavaScript

Unlike the JavaScript to Perl mappings, values are (currently¹) generally copied from Perl to JavaScript, instead of being referenced. The exceptions are Perl subs and WebPerl::JSObjects.

¹ So far, the focus of WebPerl has been to replace JavaScript with Perl, and therefore on accessing JavaScript from Perl, and not as much the other way around, that is, doing complex things with Perl from JavaScript code. For example, currently, Perl.eval() always returns a string, but could in the future be extended to return more than that, similar to WebPerl::js(), and then the passing of Perl values to JavaScript could be accomplished differently as well.

² Remember that Perl subs without an explicit return statement will implicitly return the value of the last statement (if it is an expression). This value will in turn be passed to JavaScript, which may be inefficient if this value is not needed, and it may cause errors if the return value cannot be encoded to JavaScript. Therefore it is recommended to get into the habit of adding an explicit return; at the end of subs passed to JS.


This function is a convenience function for calling JavaScript’s new. The first argument is the name of the class, the following arguments are passed to the constructor. It returns the same thing as the js() function, in this case that would be the new object. For example:

js_new('Blob', ["<html></html>"], {type=>"text/html;charset=utf-8"})

is the same as calling this in JavaScript:

new Blob(["<html></html>"], {type:"text/html;charset=utf-8"})

This function was added in WebPerl v0.05-beta.

The Mini IDE

Warning: The “mini IDE” included with WebPerl is currently not meant to be a full-featured IDE in which you’re supposed to develop your WebPerl scripts. It started out as a way to simply inspect Emscripten’s virtual filesystem, and quickly test some features. Please consider it more of a demo of some of the things that are possible, and don’t be surprised that it doesn’t have many of the features you might expect from an IDE, and it has a few “quirks” (you are of course free to provide patches, if you like ;-) ).

Note the default encoding for files is assumed to be UTF-8.

The Editor

“Run & Persist”

This controls the Perl interpreter that is part of the current page. Remember that there can be only one Perl interpreter in a page at once, and it can only run once, which means that if you edit the code in the editor, you’ll need to save the file to a location like /mnt/idb and re-load the entire page for the changes to be able to take effect.

The advantage of this mode is that it most resembles the way WebPerl is intended to be run as part of a web page - it starts once when the page is loaded (or in this case, when you click “Run”), and is then suspended after main() finishes, so that control passes back to the browser, and JavaScript can then call any handlers you’ve installed for things like click events.

“Run Multi (via IFrame)”

This provides a “hacked” way to run multiple Perl interpreters without reloading the entire page, by loading a new page in an <iframe>. The script is run there, and when main() exits, Perl is ended, and its output fetched into the “output” textarea. The advantage of this mode is that it allows quicker edit-and-run cycles, which is good for testing, the disadvantage is that you can’t really interact with Perl from the browser while it is running.

Remember to first save your script in a location like /mnt/idb, because otherwise the Perl instance in the iframe won’t be able to see it (since it gets a fresh copy of the virtual file system).

Additional notes on using WebPerl may be found in the GitHub Wiki.

Copyright (c) 2018 Hauke Daempfling (haukex@zero-g.net) at the Leibniz Institute of Freshwater Ecology and Inland Fisheries (IGB), Berlin, Germany, http://www.igb-berlin.de

Please see the “Legal” page for details.

You can find the source for this page at https://github.com/haukex/webperl/blob/gh-pages/using.md