compilation breaks on tag: 1.1.43

Erlang and Elixir XML Parsing

Fast Expat based Erlang XML parsing and manipulation library, with a strong focus on XML stream parsing from network.

It supports:

Full XML structure parsing: Suitable for small but complete XML chunks.
XML stream parsing: Suitable for large XML document, or infinite network XML stream like XMPP.

This module can parse files much faster than built-in module xmerl. Depending on file complexity and size fxml_stream:parse_element/1 can be 8-18 times faster than calling xmerl_scan:string/2.

This application was previously called p1_xml and was renamed after major optimisations to put emphasis on the fact it is damn fast.

Building

Erlang XML parser can be build as follow:

./configure && make

Erlang XML parser is a rebar-compatible OTP application. Alternatively, you can build it with rebar:

rebar compile

Dependencies

Erlang XML parser depends on Expat XML parser. You need development headers for Expat library to build it.

You can use configure options to pass custom path to Expat libraries and headers:

--with-expat=[ARG]      use Expat XML Parser from given prefix (ARG=path);
                        check standard prefixes (ARG=yes); disable (ARG=no)
--with-expat-inc=[DIR]  path to Expat XML Parser headers
--with-expat-lib=[ARG]  link options for Expat XML Parser libraries

xmlel record and types

XML elements are provided as Erlang xmlel records.

Format of the record allows defining a simple tree-like structure. xmlel record has the following fields:

name :: binary()
attrs :: [attr()]
children :: [xmlel() | cdata()]

cdata type is a tuple of the form:

{xmlcdata, CData::binary()}

attr type if a tuple of the form:

{Name::binary(), Value::binary()}

XML full structure parsing

You can definitely parse a complete XML structure with fast_xml:

$ erl -pa ebin
Erlang/OTP 17 [erts-6.3] [source] [64-bit] [smp:4:4] [async-threads:10] [hipe] [kernel-poll:false] [dtrace]

Eshell V6.3  (abort with ^G)
1> application:start(fast_xml).
ok
2> rr(fxml).
[xmlel]
3> fxml_stream:parse_element(<<"<test>content cdata</test>">>).
#xmlel{name = <<"test">>,attrs = [],
       children = [{xmlcdata,<<"content cdata">>}]}

XML Stream parsing example

You can also parse continuous stream. Our design allows decoupling very easily the process receiving the raw XML to parse from the process receiving the parsed content.

The workflow is as follow:

state = new(CallbackPID); parse(state, data); parse(state, moredata); ...

and the parsed XML fragments (stanzas) are send to CallbackPID.

With that approach you can be very flexible on how you architect your own application.

Here is an example XML stream parsing:

$ erl -pa ebin
Erlang/OTP 17 [erts-6.3] [source] [64-bit] [smp:4:4] [async-threads:10] [hipe] [kernel-poll:false] [dtrace]

Eshell V6.3  (abort with ^G)

% Start the application:
1> application:start(fast_xml).
ok

% Create a new stream, using self PID to received XML parsing event:
2> S1 = fxml_stream:new(self()).
<<>>

% Start feeding content to the XML parser.
3> S2 = fxml_stream:parse(S1, <<"<root>">>).
<<>>

% Receive Erlang message send to shell process:
4> flush().
Shell got {'$gen_event',{xmlstreamstart,<<"root">>,[]}}
ok

% Feed more content:
5> S3 = fxml_stream:parse(S2, <<"<xmlelement>content cdata</xmlelement>">>).
<<>>

% Receive more messages:
6> flush().
Shell got {'$gen_event',
              {xmlstreamelement,
                  {xmlel,<<"xmlelement">>,[],
                      [{xmlcdata,<<"content cdata">>}]}}}
ok

% Feed more content:
7> S4 = fxml_stream:parse(S3, <<"</root>">>).      
<<>>

% Receive messages:
8> flush().
Shell got {'$gen_event',{xmlstreamend,<<"root">>}}
ok

9> fxml_stream:close(S4).
true

Note how the root element is important. We expect to have the root element serve as boundary with stream start and stream end event. Then, lower level tags are passed as sub stream elements.

How does this module relate to exmpp ?

This module is a low level fast XML parser. It is not an XMPP client library like exmpp.

References

This module is use at large scale for parsing massive XML content in ejabberd XMPP server project. It is used in production in thousands of real life deployments.

Development

Test

Unit test

You can run eunit test with the command:

$ rebar eunit

Elixir / Quickcheck test

You can run test written with Elixir / Quickcheck thanks to the mix command:

MIX_EXS=test/elixir/mix.exs mix test

	static xmlns_op encode_name(state_t state, const char xml_name, ErlNifBinary *buf,
	char ns_str, char pfx_str, int top_element)
	{
	const char *parts[3];
	int i, idx = 0;

	for (i = 0; ; i++) {
	if (!xml_name[i] \|\| xml_name[i] == '\n') {
	parts[idx++] = xml_name + i;
	if (!xml_name[i])
	break;
	}
	if (idx >= 3)
	return OP_ERROR;
	}
	const char ns = NULL, name = NULL, *prefix = NULL;
	size_t ns_len = 0, name_len = 0, prefix_len = 0;

	if (idx == 1) {
	name = xml_name;
	name_len = parts[0] - xml_name;
	} else {
	ns = xml_name;
	ns_len = parts[0] - xml_name;
	name = parts[0] + 1;
	name_len = parts[1] - parts[0] - 1;
	if (idx == 3) {
	prefix = parts[1] + 1;
	prefix_len = parts[2] - parts[1] - 1;
	}
	}

processone / fast_xml Goto Github PK

fast_xml's Introduction

Erlang and Elixir XML Parsing

Building

Dependencies

xmlel record and types

XML full structure parsing

XML Stream parsing example

How does this module relate to exmpp ?

References

Development

Test

Unit test

Elixir / Quickcheck test

fast_xml's People

Contributors

Stargazers

Watchers

Forkers

fast_xml's Issues

Recommend Projects

Recommend Topics

Recommend Org