|
1
|
|
|
<?php |
|
2
|
|
|
namespace BOTK; |
|
3
|
|
|
|
|
4
|
|
|
use BOTK\Exception\FactsValidatorWarning; |
|
5
|
|
|
use BOTK\Exception\InvalidRawDataWarning; |
|
6
|
|
|
use BOTK\Exception\TooManyErrorsException; |
|
7
|
|
|
use BOTK\Exception\TooManyInsanesException; |
|
8
|
|
|
|
|
9
|
|
|
/** |
|
10
|
|
|
* Create structured data from an array of raw data (i.e. ie just a sequence of scalars) managing errors. |
|
11
|
|
|
* The class provides a RDF triple counter and tresholds for data processinge errors |
|
12
|
|
|
* Options: |
|
13
|
|
|
* 'datamapper' a callable that accepts a raw data array and create structured data as an array. Must be provided. |
|
14
|
|
|
* 'rawdataSanitizer a callable that validate raw data before datamapper. It returns an array of raw data of false if rawdata is invalid. |
|
15
|
|
|
* 'dataCleaner' a callable that filters the structured data returned by datamapper before to instanciate data model, |
|
16
|
|
|
* by default it removes all empty properties. |
|
17
|
|
|
* 'factsErrorDetector' a callable that validate computed facts. It accepts a ModelInterface and returns an error description or false. |
|
18
|
|
|
* By defaults accepted raw data that produces empty facts are considered errors. |
|
19
|
|
|
*/ |
|
20
|
|
|
class FactsFactory implements FactsFactoryInterface { |
|
21
|
|
|
|
|
22
|
|
|
protected $profile; |
|
23
|
|
|
protected $modelClass; |
|
24
|
|
|
protected $counter = array( |
|
25
|
|
|
'triple' => 0, // rdf triples in facts |
|
26
|
|
|
'error' => 0, // facts contains error |
|
27
|
|
|
'insane' => 0, // raw data unaccepted |
|
28
|
|
|
'entity' => 0, // raw data processed |
|
29
|
|
|
); |
|
30
|
|
|
|
|
31
|
|
|
|
|
32
|
1 |
|
public function __construct( array $profile =array() ) |
|
33
|
|
|
{ |
|
34
|
|
|
$defaults = array( |
|
35
|
1 |
|
'model' => 'SampleSchemaThing', |
|
36
|
|
|
'modelOptions' => array(), |
|
37
|
1 |
|
'entityThreshold' => 100, // min numbers of entity that trigger error resilence computation. |
|
38
|
1 |
|
'resilienceToErrors' => 0.3, // if more than 30% of error throws a TooManyErrorException |
|
39
|
1 |
|
'resilienceToInsanes' => 0.9, // if more than 90% of unacceptable data throws a TooManyErrorException |
|
40
|
|
|
'documentURL' => null, // used if printing metadata to specify source file URL use empty string for <> |
|
41
|
|
|
'source' => null, |
|
42
|
1 |
|
'datamapper' => function($rawdata){return $rawdata;}, |
|
43
|
1 |
|
'dataCleaner' => get_class().'::REMOVE_EMPTY', |
|
44
|
1 |
|
'factsErrorDetector' => get_class().'::NOT_EMPTY_FACTS', |
|
45
|
1 |
|
'rawdataSanitizer' => function($rawdata){return is_array($rawdata)?$rawdata:false;}, |
|
46
|
|
|
); |
|
47
|
1 |
|
$this->profile = array_merge($defaults,$profile); |
|
48
|
1 |
|
$this->modelClass = class_exists($this->profile['model']) |
|
49
|
|
|
?$this->profile['model'] |
|
50
|
1 |
|
:('\BOTK\Model\\'.$this->profile['model']); |
|
51
|
|
|
|
|
52
|
1 |
|
if( !class_exists($this->modelClass) || !is_subclass_of($this->modelClass, '\BOTK\ModelInterface')){ |
|
53
|
|
|
throw new \InvalidArgumentException("The provided model ({$this->profile['model']} is unknown"); |
|
54
|
|
|
} |
|
55
|
1 |
|
if( !is_callable($this->profile['datamapper'])) { |
|
56
|
|
|
throw new \InvalidArgumentException("Invalid datamapper callback"); |
|
57
|
|
|
} |
|
58
|
1 |
|
if( !is_callable($this->profile['dataCleaner'])) { |
|
59
|
|
|
throw new \InvalidArgumentException("Invalid dataCleaner callback"); |
|
60
|
|
|
} |
|
61
|
1 |
|
if( !is_callable($this->profile['rawdataSanitizer'])) { |
|
62
|
|
|
throw new \InvalidArgumentException("Invalid rawdataSanitizer callback"); |
|
63
|
|
|
} |
|
64
|
1 |
|
if( !is_callable($this->profile['factsErrorDetector'])) { |
|
65
|
|
|
throw new \InvalidArgumentException("Invalid factsErrorDetector callback"); |
|
66
|
|
|
} |
|
67
|
1 |
|
} |
|
68
|
|
|
|
|
69
|
|
|
|
|
70
|
|
|
/** |
|
71
|
|
|
* two level filter array, a default for dataCleaner callback |
|
72
|
|
|
*/ |
|
73
|
1 |
|
public static function REMOVE_EMPTY( array $data) |
|
74
|
|
|
{ |
|
75
|
1 |
|
$a = array(); |
|
76
|
1 |
|
foreach ($data as $key => $value) { |
|
77
|
1 |
|
$a[$key] = is_array($value)?array_filter($value):$value; |
|
78
|
|
|
} |
|
79
|
1 |
|
return array_filter($a); |
|
80
|
|
|
} |
|
81
|
|
|
|
|
82
|
|
|
|
|
83
|
|
|
/** |
|
84
|
|
|
* a default for dataValidator callback |
|
85
|
1 |
|
*/ |
|
86
|
|
|
public static function NOT_EMPTY_FACTS( \BOTK\ModelInterface $data) |
|
87
|
1 |
|
{ |
|
88
|
|
|
return $data->getTripleCount()?false:'No facts found.'; |
|
89
|
|
|
} |
|
90
|
|
|
|
|
91
|
|
|
|
|
92
|
|
|
/** |
|
93
|
|
|
* create facts from rawdata. Please nothe that null facts does not means always an error (i.e. no facts is a fact). |
|
94
|
|
|
* if you do not want empty facts use dataValidator |
|
95
|
1 |
|
*/ |
|
96
|
|
|
public function factualize($rawData, object $globalStorage=null) |
|
97
|
1 |
|
{ |
|
98
|
1 |
|
|
|
99
|
1 |
|
$rawdataSanitizer = $this->profile['rawdataSanitizer']; |
|
100
|
|
|
$validRawData = $rawdataSanitizer($rawData); |
|
101
|
1 |
|
$this->counter['entity']++; |
|
102
|
1 |
|
|
|
103
|
1 |
|
if (!empty($validRawData)){ |
|
104
|
1 |
|
$datamapper = $this->profile['datamapper']; |
|
105
|
1 |
|
$dataCleaner = $this->profile['dataCleaner']; |
|
106
|
1 |
|
$factsErrorDetector = $this->profile['factsErrorDetector']; |
|
107
|
1 |
|
$data =$dataCleaner($datamapper($validRawData)); |
|
108
|
1 |
|
$facts = call_user_func($this->modelClass.'::fromArray', $data, $this->profile['modelOptions'], $globalStorage); |
|
109
|
|
|
$this->counter['triple'] += $facts->getTripleCount(); |
|
110
|
1 |
|
if($error=$factsErrorDetector($facts)){ |
|
111
|
|
|
$this->counter['error']++; |
|
112
|
|
|
throw new FactsValidatorWarning($error,$facts); |
|
113
|
|
|
} |
|
114
|
|
|
} else { |
|
115
|
|
|
$this->counter['insane']++; |
|
116
|
|
|
throw new InvalidRawDataWarning("Invalid rawdata",$rawData); |
|
117
|
|
|
} |
|
118
|
1 |
|
|
|
119
|
1 |
|
// ensure that not too many errors |
|
120
|
1 |
|
$errorRate = ($this->counter['error']/$this->counter['entity']); |
|
121
|
|
|
if(( $this->counter['entity'] > $this->profile['entityThreshold']) |
|
122
|
|
|
&& ( $errorRate > $this->profile['resilienceToErrors'])){ |
|
123
|
|
|
$x = $this->profile['resilienceToErrors']*100; |
|
124
|
|
|
throw new TooManyErrorsException("Error rate in data processing exceeded the $x% threshold"); |
|
125
|
|
|
} |
|
126
|
1 |
|
|
|
127
|
1 |
|
// ensure that not too many insaness raw data |
|
128
|
1 |
|
$insaneRate = ($this->counter['insane']/$this->counter['entity']); |
|
129
|
|
|
if(( $this->counter['entity'] > $this->profile['entityThreshold']) |
|
130
|
|
|
&& ($insaneRate > $this->profile['resilienceToInsanes'])){ |
|
131
|
|
|
$x = $this->profile['resilienceToInsanes']*100; |
|
132
|
|
|
throw new TooManyInsanesException("Unacceptable data rate exceeded the $x% threshold"); |
|
133
|
1 |
|
} |
|
134
|
|
|
|
|
135
|
|
|
return $facts; |
|
136
|
|
|
} |
|
137
|
|
|
|
|
138
|
|
|
|
|
139
|
|
|
public function generateLinkedDataHeader() |
|
140
|
|
|
{ |
|
141
|
|
|
$metadata=''; |
|
142
|
|
|
if(!is_null($this->profile['documentURL'])) { |
|
143
|
|
|
// Requires foaf:, dct: and void: prefixes to be defined in model |
|
144
|
|
|
$metadata.= "\n<{$this->profile['documentURL']}> a foaf:Document .\n"; |
|
145
|
|
|
if(!empty($this->profile['source'])){ |
|
146
|
|
|
$metadata .= "<{$this->profile['documentURL']}> dct:source <{$this->profile['source']}>. \n"; |
|
147
|
|
|
} |
|
148
|
|
|
$metadata .= "[] a void:Dataset; dct:format \"text/turtle\" ; void:datadump <{$this->profile['documentURL']}>;void:triples {$this->counter['triple']} ;void:entities {$this->counter['entity']}.\n"; |
|
149
|
|
|
} |
|
150
|
|
|
$base = $this->profile['modelOptions']['base']['default']?? null; |
|
151
|
|
|
return call_user_func($this->modelClass.'::getTurtleHeader',$base) . $metadata; |
|
152
|
|
|
} |
|
153
|
|
|
|
|
154
|
|
|
|
|
155
|
|
|
public function generateLinkedDataFooter() |
|
156
|
|
|
{ |
|
157
|
|
|
return "#Generated {$this->counter['triple']} good triples from {$this->counter['entity']} entities ({$this->counter['insane']} ignored), {$this->counter['error']} errors\n"; |
|
158
|
|
|
} |
|
159
|
|
|
|
|
160
|
|
|
|
|
161
|
|
|
public function addToCounter($counter,$val=1) |
|
162
|
|
|
{ |
|
163
|
|
|
if(!array_key_exists($counter,$this->counter)){ |
|
164
|
|
|
throw new \InvalidArgumentException("Invalid counter name"); |
|
165
|
|
|
} |
|
166
|
|
|
$this->counter[$counter]+= intval($val); |
|
167
|
|
|
} |
|
168
|
|
|
|
|
169
|
|
|
|
|
170
|
|
|
public function getCounters() |
|
171
|
|
|
{ |
|
172
|
|
|
return $this->counter; |
|
173
|
|
|
} |
|
174
|
|
|
|
|
175
|
|
|
} |
linkeddatacenter /
BOTK-core
