private RegexRecognizer CreateUriRecognizer(System.Globalization.CultureInfo actualCulture, int priority)
{
// TODO this shouldn't be a recognizer but rather a simple classifier after the
// default fallback recognizer (for performance reasons)
// TODO set context restrictions of the recognizer
RegexRecognizer recog = new RegexRecognizer(TokenType.Uri, priority, "URI", "DEFAULT_URI_REGOCNIZER");
CharacterSet first = new CharacterSet();
// http, https, mailto, ftp, file
first.Add('h');
first.Add('H');
first.Add('m');
first.Add('M');
first.Add('f');
first.Add('F');
recog.Add("(mailto:|((https|http|ftp|file)://))[\\p{L}\\p{N}\\p{Pc}\\p{Pd}\\p{Po}\\p{S}-['\"<>]]*[\\p{L}\\p{N}\\p{Pc}\\p{Pd}\\p{S}/]", first, true);
// not sure about this one:
recog.OnlyIfFollowedByNonwordCharacter
= Core.CultureInfoExtensions.UseBlankAsWordSeparator(actualCulture);
return(recog);
}
private RegexRecognizer CreateVariableRecognizer(Core.Resources.IResourceDataAccessor accessor,
System.Globalization.CultureInfo actualCulture)
{
Wordlist wl = new Wordlist();
// TODO also create the recognizer if no variables are defined/available?
using (System.IO.Stream data = accessor.ReadResourceData(_Culture, Core.Resources.LanguageResourceType.Variables, true))
{
if (data != null)
{
wl.Load(data);
}
}
if (wl.Count == 0)
{
return(null);
}
// TODO set context restrictions of the recognizer
RegexRecognizer recog = new RegexRecognizer(TokenType.Variable, 100, "VAR", "DEFAULT_VAR_REGOCNIZER");
CharacterSet first;
string rx = wl.GetRegularExpression(out first);
recog.Add(rx, first);
recog.OnlyIfFollowedByNonwordCharacter =
Core.CultureInfoExtensions.UseBlankAsWordSeparator(actualCulture);
return(recog);
}
private Recognizer CreateHeadingNumberRecognizer(System.Globalization.CultureInfo culture, int priority)
{
try
{
// TODO use culture's digits but don't use Regex \d placeholder as it's not culture sensitive
// TODO set context (word boundaries)
// TODO make IP address an alphanumeric token with placeable features?
// TODO treat all alphanumeric tokens as placeables?
string pattern = "[0-9]+(\\.([0-9]{1,3}))+";
CharacterSet first = new CharacterSet();
first.Add('0', '9');
RegexRecognizer recog = new RegexRecognizer(TokenType.OtherTextPlaceable,
priority, "GENHNUMPLC", "GENERIC_HEADINGNUMBER_RECOGNIZER", true);
recog.Add(pattern, first);
// TODO is this culture-dependent?
recog.OnlyIfFollowedByNonwordCharacter = true;
return(recog);
}
catch // (System.Exception e)
{
return(null);
}
}
private RegexRecognizer CreateAcronymRecognizer(System.Globalization.CultureInfo actualCulture, int priority)
{
// TODO this shouldn't be a recognizer but rather a simple classifier after the
// default fallback recognizer (for performance reasons)
// TODO set context restrictions of the recognizer
// TODO acronyms shouldn't be recognized in an all-caps context.
// TODO we may also include some additional special symbols such as "_" "."
RegexRecognizer recog = new RegexRecognizer(TokenType.Acronym, priority, "ACR", "DEFAULT_ACR_REGOCNIZER");
CharacterSet first = new CharacterSet();
first.Add(System.Globalization.UnicodeCategory.UppercaseLetter);
// TODO doesn't catch e.g. OePNV, BfA, APIs
// recog.Add("[A-Z][A-Z&]{0,2}[A-Z]{1,3}", first);
recog.Add(@"\p{Lu}[\p{Lu}&]{0,4}\p{Lu}", first);
recog.OnlyIfFollowedByNonwordCharacter
= Core.CultureInfoExtensions.UseBlankAsWordSeparator(actualCulture);
return(recog);
}
/// <summary>
/// Initialize the tokenizer parameters from the tokenizer setup information.
/// The resource data accessor is only used for retrieving variable values.
/// </summary>
/// <param name="setup">The tokenizer setup to use</param>
/// <param name="accessor">A resource data accessor, to initialize the variables list</param>
public TokenizerParameters(TokenizerSetup setup, Core.Resources.IResourceDataAccessor accessor)
{
if (setup == null)
{
throw new ArgumentNullException("setup");
}
if (accessor == null)
{
accessor = new ResourceFileResourceAccessor();
}
_BreakOnWhitespace = setup.BreakOnWhitespace;
_CreateWhitespaceTokens = setup.CreateWhitespaceTokens;
_Culture = Core.CultureInfoExtensions.GetCultureInfo(setup.CultureName);
_Recognizers = new List <Recognizer>();
_ReclassifyAcronyms = false;
// we need to determine a region-qualified culture since neutral cultures
// don't have date/time/number pattern info
System.Globalization.CultureInfo actualCulture = _Culture;
if (_Culture.IsNeutralCulture)
{
actualCulture = Core.CultureInfoExtensions.GetRegionQualifiedCulture(_Culture);
}
if ((setup.BuiltinRecognizers & Core.Tokenization.BuiltinRecognizers.RecognizeDates) != 0)
{
AddRecognizer(DateTimeRecognizer.Create(accessor, actualCulture,
DateTimePatternType.ShortDate | DateTimePatternType.LongDate, 100));
}
if ((setup.BuiltinRecognizers & Core.Tokenization.BuiltinRecognizers.RecognizeTimes) != 0)
{
AddRecognizer(DateTimeRecognizer.Create(accessor, actualCulture,
DateTimePatternType.ShortTime | DateTimePatternType.LongTime, 100));
}
if ((setup.BuiltinRecognizers & Core.Tokenization.BuiltinRecognizers.RecognizeNumbers) != 0)
{
if (UseRXNumberRecognizer)
{
AddRecognizer(NumberRegexRecognizer.Create(actualCulture, 100));
}
else
{
AddRecognizer(NumberFSTRecognizer.Create(accessor, actualCulture, 100));
}
// it does not make sense to recognize ordinal numbers if they don't become placeables and
// don't participate in auto-localization. They'd also need to be auto-localized (3. -> 3rd)
bool createOrdinalNumberRecognizer = false;
if (createOrdinalNumberRecognizer)
{
// add special recognizer for ordinal numbers if ordinal followers are available
// [0-9]+\. (?=[OrdinalFollowers])
if (accessor.GetResourceStatus(_Culture, Core.Resources.LanguageResourceType.OrdinalFollowers, true) != Core.Resources.ResourceStatus.NotAvailable)
{
Wordlist ordinalFollowers = new Wordlist();
CharacterSet dummy;
ordinalFollowers.Load(accessor.ReadResourceData(_Culture, Core.Resources.LanguageResourceType.OrdinalFollowers, true));
if (ordinalFollowers.Count > 0)
{
string ordinalNumbersRX = "[0-9]+\\.(?=[ \u00A0]" + ordinalFollowers.GetRegularExpression(out dummy) + "\\s)";
RegexRecognizer ordinalNumbersRecognizer = new RegexRecognizer(TokenType.Word, 100, "ORDINALNUMBER", "Ordinal Number Recognizer");
CharacterSet ordinalFirst = new CharacterSet();
ordinalFirst.Add('0', '9');
ordinalNumbersRecognizer.Add(ordinalNumbersRX, ordinalFirst);
AddRecognizer(ordinalNumbersRecognizer);
}
}
}
}
else
{
// TODO should we still add a rudimentary recognizer for alpha-numerals?
}
// TODO other recognizer types (for builtin token classes)
if ((setup.BuiltinRecognizers & Core.Tokenization.BuiltinRecognizers.RecognizeAcronyms) != 0)
{
RegexRecognizer recog = CreateAcronymRecognizer(actualCulture, 100);
if (recog != null)
{
_ReclassifyAcronyms = true;
AddRecognizer(recog);
}
// this shouldn't be in the "acronym" setting but it's too late for a UI change...
recog = CreateUriRecognizer(actualCulture, 100);
AddRecognizer(recog);
// TODO make IP address recognizer optional?
AddRecognizer(CreateIPAddressRecognizer(actualCulture, 101));
// AddRecognizer(CreateHeadingNumberRecognizer(actualCulture, 50));
}
if ((setup.BuiltinRecognizers & Core.Tokenization.BuiltinRecognizers.RecognizeVariables) != 0)
{
if (accessor != null)
{
try
{
RegexRecognizer recog = CreateVariableRecognizer(accessor, actualCulture);
if (recog != null)
{
AddRecognizer(recog);
}
}
catch // (System.Exception e)
{
// nop - ignore errors
}
}
}
if ((setup.BuiltinRecognizers & Core.Tokenization.BuiltinRecognizers.RecognizeMeasurements) != 0)
{
Recognizer recog;
if (UseRXMeasurementRecognizer)
{
recog = MeasureRegexRecognizer.Create(actualCulture, 100);
}
else
{
recog = MeasureFSTRecognizer.Create(accessor, actualCulture, 100);
}
AddRecognizer(recog);
// disable for the time being due to performance issues
//if (accessor.GetResourceStatus(actualCulture, Core.Resources.LanguageResourceType.CurrencySymbols, true) != Core.Resources.ResourceStatus.NotAvailable)
//{
// recog = CreateCurrencyRecognizer(accessor, actualCulture);
// AddRecognizer(recog);
//}
}
#if false
// TODO NOTE this slows down the performance too much - need to find a better way
if (accessor.GetResourceStatus(actualCulture, Core.Resources.LanguageResourceType.Abbreviations, true) != Core.Resources.ResourceStatus.NotAvailable)
{
// add an abbreviation recognizer
Wordlist abbreviations = new Wordlist();
CharacterSet first;
abbreviations.Load(accessor.ReadResourceData(actualCulture, Core.Resources.LanguageResourceType.Abbreviations, true));
string abbreviationsRX = abbreviations.GetRegularExpression(out first) + @"(?=\W)";
RegexRecognizer abbreviationsRecognizer = new RegexRecognizer(TokenType.Abbreviation, 101, "ABBREVIATION", "Abbreviation Recognizer");
abbreviationsRecognizer.Add(abbreviationsRX, first);
AddRecognizer(abbreviationsRecognizer);
}
#endif
{
Recognizer recog;
bool split = setup.SeparateClitics && Core.CultureInfoExtensions.UsesClitics(_Culture);
recog = CreateDefaultFallbackRecognizer(split, accessor);
AddRecognizer(recog);
}
SortRecognizers();
}