Class: CharlockHolmes::EncodingDetector

Inherits:
Object
  • Object
show all
Defined in:
lib/charlock_holmes/encoding_detector.rb,
ext/charlock_holmes/encoding_detector.c

Constant Summary collapse

DEFAULT_BINARY_SCAN_LEN =

Default length for which to scan content for NULL bytes

1024*1024
BINARY = 
'binary'

Instance Attribute Summary collapse

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(scan_len = DEFAULT_BINARY_SCAN_LEN) ⇒ EncodingDetector

Returns a new instance of EncodingDetector.



11
12
13
 
# File 'lib/charlock_holmes/encoding_detector.rb', line 11

def initialize(scan_len=DEFAULT_BINARY_SCAN_LEN)
  @binary_scan_length = scan_len
end

Instance Attribute Details

#binary_scan_length ⇒ Object

Length for which to scan content for NULL bytes



7
8
9
 
# File 'lib/charlock_holmes/encoding_detector.rb', line 7

def binary_scan_length
  @binary_scan_length
end

Class Method Details

.build_encoding_table ⇒ Object

Builds the ENCODING_TABLE hash by running through the list of supported encodings in the ICU detection API and trying to map them to supported encodings in Ruby. This is built dynamically so as to take advantage of ICU upgrades which may have support for more encodings in the future.

Returns nothing.



65
66
67
68
69
70
71
72
73
 
# File 'lib/charlock_holmes/encoding_detector.rb', line 65

def self.build_encoding_table
  supported_encodings.each do |name|
    @encoding_table[name] = begin
      ::Encoding.find(name).name
    rescue ArgumentError
      BINARY
    end
  end
end

.detect(str, hint_enc = nil) ⇒ Object

Attempt to detect the encoding of this string

NOTE: This will create a new CharlockHolmes::EncodingDetector instance on every call as well as use the default binary scan length

str - a String, what you want to detect the encoding of hint_enc - an optional String (like “UTF-8”), the encoding name which will

be used as an additional hint to the charset detector

Returns: a Hash with :encoding, :language, :type and :confidence



25
26
27
 
# File 'lib/charlock_holmes/encoding_detector.rb', line 25

def self.detect(str, hint_enc=nil)
  new.detect(str, hint_enc)
end

.detect_all(str, hint_enc = nil) ⇒ Object

Attempt to detect the encoding of this string, and return a list with all the possible encodings that match it.

NOTE: This will create a new CharlockHolmes::EncodingDetector instance on every call as well as use the default binary scan length

str - a String, what you want to detect the encoding of hint_enc - an optional String (like “UTF-8”), the encoding name which will

be used as an additional hint to the charset detector

Returns: an Array with zero or more Hashes, each one of them with with :encoding, :language, :type and :confidence



41
42
43
 
# File 'lib/charlock_holmes/encoding_detector.rb', line 41

def self.detect_all(str, hint_enc=nil)
  new.detect_all(str, hint_enc)
end

.encoding_table ⇒ Object 



53
54
55
 
# File 'lib/charlock_holmes/encoding_detector.rb', line 53

def self.encoding_table
  @encoding_table
end

.detectable_encodings=(EncodingDetector) ⇒ Object

The list of detectable encodings supported by this library

Returns: an Array of Strings



298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
 
# File 'ext/charlock_holmes/encoding_detector.c', line 298

static VALUE rb_get_supported_encodings(VALUE klass)
{
	UCharsetDetector *csd;
	UErrorCode status = U_ZERO_ERROR;
	UEnumeration *encoding_list;
	VALUE rb_encoding_list;
	int32_t enc_count;
	int32_t i;
	const char *enc_name;
	int32_t enc_name_len;

	rb_encoding_list = rb_iv_get(klass, "encoding_list");

	// lazily populate the list
	if (NIL_P(rb_encoding_list)) {
		csd = ucsdet_open(&status);

		encoding_list = ucsdet_getAllDetectableCharsets(csd, &status);
		rb_encoding_list = rb_ary_new();
		enc_count = uenum_count(encoding_list, &status);

		rb_ary_push(rb_encoding_list, charlock_new_str2("windows-1250"));
		rb_ary_push(rb_encoding_list, charlock_new_str2("windows-1252"));
		rb_ary_push(rb_encoding_list, charlock_new_str2("windows-1253"));
		rb_ary_push(rb_encoding_list, charlock_new_str2("windows-1254"));
		rb_ary_push(rb_encoding_list, charlock_new_str2("windows-1255"));

		for(i=0; i < enc_count; i++) {
			enc_name = uenum_next(encoding_list, &enc_name_len, &status);
			rb_ary_push(rb_encoding_list, charlock_new_str(enc_name, enc_name_len));
		}

		rb_iv_set(klass, "encoding_list", rb_encoding_list);
		ucsdet_close(csd);
	}

	return rb_encoding_list;
}

Instance Method Details

#detection_hash=(EncodingDetector) ⇒ Object

Attempt to detect the encoding of this string

str - a String, what you want to detect the encoding of hint_enc - an optional String (like “UTF-8”), the encoding name which will

be used as an additional hint to the charset detector

Returns: a Hash with :encoding, :language, :type and :confidence



160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
 
# File 'ext/charlock_holmes/encoding_detector.c', line 160

static VALUE rb_encdec_detect(int argc, VALUE *argv, VALUE self)
{
	UErrorCode status = U_ZERO_ERROR;
	charlock_detector_t *detector;
	VALUE rb_str;
	VALUE rb_enc_hint;

	rb_scan_args(argc, argv, "11", &rb_str, &rb_enc_hint);

	Check_Type(rb_str, T_STRING);
	Data_Get_Struct(self, charlock_detector_t, detector);

	// first lets see if this is binary content
	if (detect_binary_content(self, rb_str)) {
		return rb_encdec_binarymatch();
	}

	// if we got here - the data doesn't look like binary
	// lets try to figure out what encoding the text is in
	ucsdet_setText(detector->csd, RSTRING_PTR(rb_str), (int32_t)RSTRING_LEN(rb_str), &status);

	if (!NIL_P(rb_enc_hint)) {
		Check_Type(rb_enc_hint, T_STRING);
		ucsdet_setDeclaredEncoding(detector->csd, RSTRING_PTR(rb_enc_hint), RSTRING_LEN(rb_enc_hint), &status);
	}

	return rb_encdec_buildmatch(ucsdet_detect(detector->csd, &status));
}

#detection_hash_array=(EncodingDetector) ⇒ Object

Attempt to detect the encoding of this string, and return a list with all the possible encodings that match it.

str - a String, what you want to detect the encoding of hint_enc - an optional String (like “UTF-8”), the encoding name which will

be used as an additional hint to the charset detector

Returns: an Array with zero or more Hashes,

each one of them with with :encoding, :language, :type and :confidence



204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
 
# File 'ext/charlock_holmes/encoding_detector.c', line 204

static VALUE rb_encdec_detect_all(int argc, VALUE *argv, VALUE self)
{
	UErrorCode status = U_ZERO_ERROR;
	charlock_detector_t *detector;
	const UCharsetMatch **csm;
	VALUE rb_ret;
	int i, match_count;
	VALUE rb_str;
	VALUE rb_enc_hint;
	VALUE binary_match;

	rb_scan_args(argc, argv, "11", &rb_str, &rb_enc_hint);

	Check_Type(rb_str, T_STRING);
	Data_Get_Struct(self, charlock_detector_t, detector);

	rb_ret = rb_ary_new();

	// first lets see if this is binary content
	binary_match = Qnil;
	if (detect_binary_content(self, rb_str)) {
		binary_match = rb_encdec_binarymatch();
	}

	ucsdet_setText(detector->csd, RSTRING_PTR(rb_str), (int32_t)RSTRING_LEN(rb_str), &status);

	if (!NIL_P(rb_enc_hint)) {
		Check_Type(rb_enc_hint, T_STRING);
		ucsdet_setDeclaredEncoding(detector->csd, RSTRING_PTR(rb_enc_hint), RSTRING_LEN(rb_enc_hint), &status);
	}

	csm = ucsdet_detectAll(detector->csd, &match_count, &status);

	for (i = 0; i < match_count; ++i) {
		rb_ary_push(rb_ret, rb_encdec_buildmatch(csm[i]));
	}

	if (!NIL_P(binary_match))
		rb_ary_unshift(rb_ret, binary_match);

	return rb_ret;
}

#true ⇒ Boolean

Attempt to detect if a string is binary or text

str - a String, what you want to perform the binary check on

Returns: true or false

Returns:

  • (Boolean) 


141
142
143
144
145
146
147
 
# File 'ext/charlock_holmes/encoding_detector.c', line 141

static VALUE rb_encdec_is_binary(VALUE self, VALUE str)
{
	if (detect_binary_content(self, str))
		return Qtrue;
	else
		return Qfalse;
}

#strip_tags? ⇒ Object Also known as: strip_tags?

Returns whether or not the strip_tags flag is set on this detector

Returns: Boolean



254
255
256
257
258
259
260
261
262
263
264
265
266
267
 
# File 'ext/charlock_holmes/encoding_detector.c', line 254

static VALUE rb_get_strip_tags(VALUE self)
{
	charlock_detector_t *detector;
	UBool val;
	VALUE rb_val;

	Data_Get_Struct(self, charlock_detector_t, detector);

	val = ucsdet_isInputFilterEnabled(detector->csd);

	rb_val = val == 1 ? Qtrue : Qfalse;

	return rb_val;
}

#strip_tags=(true) ⇒ Object

Enable or disable the stripping of HTML/XML tags from the input before attempting any detection

Returns: Boolean, the value passed



277
278
279
280
281
282
283
284
285
286
287
288
289
 
# File 'ext/charlock_holmes/encoding_detector.c', line 277

static VALUE rb_set_strip_tags(VALUE self, VALUE rb_val)
{
	charlock_detector_t *detector;
	UBool val;

	Data_Get_Struct(self, charlock_detector_t, detector);

	val = rb_val == Qtrue ? 1 : 0;

	ucsdet_enableInputFilter(detector->csd, val);

	return rb_val;
}