Richard Jones, Esq. » php http://www.metabrew.com Erlang, PHP, C, C++, Java, PostgreSQL, MySQL, Hadoop, Linux, awk, bash, sed, grep, screen, vim, irc, ssh etc... Sun, 20 Dec 2009 18:59:32 +0000 en hourly 1 http://wordpress.org/?v=3.0 Reading Serialized PHP Objects from Erlang http://www.metabrew.com/article/reading-serialized-php-objects-from-erlang/ http://www.metabrew.com/article/reading-serialized-php-objects-from-erlang/#comments Sat, 27 Sep 2008 22:42:20 +0000 RJ http://www.metabrew.com/wordpress/?p=15 I started writing some Erlang recently. The vast majority of data I need to access from Erlang resides in cached, serialized php objects. Here’s what I came up with to turn a serialized php object into a sort of nested Erlang proplist thing.

  1. <?php
  2. $s = array(123, ‘hello’, 3.14, array(‘a’=>‘foo’, ‘b’=>‘bar’));
  3. print serialize($s);
  4. ?>

This gives:

a:4:{i:0;i:123;i:1;s:5:"hello";i:2;d:3.14;i:3;a:2:{s:1:"a";s:3:"foo";s:1:"b";s:3:"bar";}}

It’s not hard to see how the (relatively undocumented) PHP serialization format works. Here’s what it becomes in Erlang:


1> php:unserialize("a:4:{i:0;i:123;i:1;s:5:\"hello\";i:2;d:3.14;i:3;a:2:{s:1:\"a\";s:3:\"foo\";s:1:\"b\";s:3:\"bar\";}}").

{[[{0,123},
{1,<<"hello">>},
{2,3.14},
{3,[{a,<<"foo">>},{b,<<"bar">>}]}]],
[]}

Here’s what it does with objects:

  1. <?php
  2. class ExampleClass {
  3.     var $id = 123;
  4.     var $name = "RJ";
  5.     var $languages = array(‘php’, ‘erlang’, ‘etc’);
  6. }
  7. $s = new ExampleClass();
  8. $ser = serialize($s);
  9. print serialize($str);
  10. ?>


2> php:unserialize("O:12:\"ExampleClass\":3:{s:2:\"id\";i:123;s:4:\"name\";s:2:\"RJ\";s:9:\"languages\";a:3:{i:0;s:3:\"php\";i:1;s:6:\"erlang\";i:2;s:3:\"etc\";}}").
{[{class,"ExampleClass",
[{id,123},
{name,<<"RJ">>},
{languages,[{0,<<"php">>},
{1,<<"erlang">>},
{2,<<"etc">>}]}]}],
[]}

Due to a combination of PHP’s “relaxed” type system, an old database abstraction library, and munging things in and out of memcached, we sometimes end up with numeric properties, such as ‘id’, represented as strings by PHP. To mitigate this, I ended up with some nasty code that forces certain properties to a predefined type (“id” is always an int, etc..). Yuk. Anyway, here’s the Erlang module:

  1. %
  2. % Takes a serialized php object and turns it into an erlang data structure
  3. %
  4. -module(php).
  5. -author(‘Richard Jones <rj at last.fm>’).
  6. -export([unserialize/1]).
  7.  
  8. % Usage:  {Result, Leftover} = php:unserialize(…)
  9.  
  10. unserialize(S) when is_binary(S)    -> unserialize(binary_to_list(S));
  11. unserialize(S) when is_list(S)      -> takeval(S, 1).
  12.  
  13. % Internal stuff
  14.  
  15. takeval(Str, Num) ->
  16.     {Parsed, Remains} = takeval(Str, Num, []),
  17.     { lists:reverse(Parsed), Remains }.
  18.  
  19. takeval([$} | Leftover], 0, Acc)    -> {Acc, Leftover};
  20. takeval(Str, 0, Acc)                -> {Acc, Str};
  21. takeval([], 0, Acc)                 -> Acc;
  22.  
  23. takeval(Str, Num, Acc) ->
  24.     {Val, Rest} = phpval(Str),
  25.     %Lots of tracing if you enable this:
  26.     %io:format("\nState\n Str: ~s\n Num: ~w\n Acc:~w\n", [Str,Num,Acc]),
  27.     %io:format("-Val: ~w\n-Rest: ~s\n\n",[Val, Rest]),
  28.     takeval(Rest, Num-1, [Val | Acc]).
  29.  
  30. %
  31. % Parse induvidual php values.
  32. % a "phpval" here is T:val; where T is the type code for int, object, array etc..
  33. %
  34.  
  35. % Simple ones:
  36. phpval([])                      -> [];
  37. phpval([ $} | Rest ])           -> phpval(Rest);    % skip }
  38. phpval([$N,$;|Rest])            -> {null, Rest};    % null
  39. phpval([$b,$:,$1,$; | Rest])    -> {true, Rest};    % true
  40. phpval([$b,$:,$0,$; | Rest])    -> {false, Rest};   % false
  41.  
  42. % r seems to be a recursive reference to something, represented as an int.
  43. phpval([$r, $: | Rest]) ->
  44.     {RefNum, [$; | Rest1]} = string:to_integer(Rest),
  45.     {{php_ref, RefNum}, Rest1};
  46.  
  47. % int
  48. phpval([$i, $: | Rest])->
  49.     {Num, [$; | Rest1]} = string:to_integer(Rest),
  50.     {Num, Rest1};
  51.  
  52. % double / float
  53. % NB: php floats can be ints, and string:to_float doesn’t like that.
  54. phpval(X=[$d, $: | Rest]) ->
  55.     {Num, [$; | Rest1]} = case string:to_float(Rest) of
  56.                             {error, no_float} -> string:to_integer(Rest);
  57.                             {N,R} -> {N,R}
  58.     end,
  59.     {Num, Rest1};
  60.  
  61. % string
  62. phpval([$s, $: | Rest]) ->
  63.     {Len, [$: | Rest1]} =string:to_integer(Rest),
  64.     S = list_to_binary(string:sub_string(Rest1, 2, Len+1)),
  65.     {S, lists:nthtail(Len+3, Rest1)};
  66.  
  67. % array
  68. phpval([$a, $: | Rest]) ->
  69.     {NumEntries, [$:, ${ | Rest1]} =string:to_integer(Rest),
  70.     {Array, Rest2} = takeval(Rest1, NumEntries*2),
  71.     {arraytidy(Array), Rest2};
  72.  
  73. % object O:4:\"User\":53:{
  74. phpval([$O, $: | Rest]) ->
  75.     {ClassnameLen, [$: | Rest1]} =string:to_integer(Rest),
  76.     % Rest1: "classname":NumEnt:{..
  77.     Classname = string:sub_string(Rest1, 2, ClassnameLen+1),
  78.     Rest1b = lists:nthtail(ClassnameLen+3, Rest1),
  79.     {NumEntries, [$:, ${ | Rest2]} = string:to_integer(Rest1b),
  80.     {Classvals, Rest3} = takeval(Rest2, NumEntries*2),
  81.     {{class, Classname, arraytidy(Classvals)}, Rest3}.
  82.  
  83. %%
  84. %% Helpers:
  85. %%
  86.  
  87. % convert [ k1,v1,k2,v2,k3,v3 ] into [ {k1,v2}, {k2,v2}, {k3,v3} ]
  88. arraytidy(L) ->
  89.     lists:reverse(lists:foldl(fun arraytidy/2, [], L)).
  90.  
  91. arraytidy(El, [{key___partial, K} | L]) -> [{atomize(K), El} | L];
  92.  
  93. arraytidy(El, L) -> [{key___partial, El} | L].
  94.  
  95. %% Make properties or keys into atoms
  96. atomize(K) when is_binary(K) ->
  97.     atomize(binary_to_list(K));
  98. atomize(K) when is_list(K) ->
  99.     list_to_atom(string:to_lower(K));
  100. atomize(K) -> K.
]]> http://www.metabrew.com/article/reading-serialized-php-objects-from-erlang//feed 9