NAME String::Tagged - string buffers with value tags on extents SYNOPSIS use String::Tagged; my $st = String::Tagged->new( "An important message" ); $st->apply_tag( 3, 9, bold => 1 ); $st->iter_substr_nooverlap( sub { my ( $substring, %tags ) = @_; print $tags{bold} ? "<b>$substring</b>" : $substring; } ); DESCRIPTION This module implements an object class, instances of which store a (mutable) string buffer that supports tags. A tag is a name/value pair that applies to some extent of the underlying string. The types of tag names ought to be strings, or at least values that are well-behaved as strings, as the names will often be used as the keys in hashes or applied to the eq operator. The types of tag values are not restricted - any scalar will do. This could be a simple integer or string, ARRAY or HASH reference, or even a CODE reference containing an event handler of some kind. Tags may be arbitrarily overlapped. Any given offset within the string has in effect, a set of uniquely named tags. Tags of different names are independent. For tags of the same name, only the latest, shortest tag takes effect. For example, consider a string with three tags represented here: Here is my string with tags [-------------------------] foo => 1 [-------] foo => 2 [---] bar => 3 Every character in this string has a tag named foo. The value of this tag is 2 for the words my and string and the space inbetween, and 1 elsewhere. Additionally, the words is and my and the space between them also have the tag bar with a value 3. Since String::Tagged does not understand the significance of the tag values it therefore cannot detect if two neighbouring tags really contain the same semantic idea. Consider the following string: A string with words [-------] type => "message" [--------] type => "message" This string contains two tags. String::Tagged will treat this as two different tag values as far as iter_tags_nooverlap is concerned, even though get_tag_at yields the same value for the type tag at any position in the string. The merge_tags method may be used to merge tag extents of tags that should be considered as equal. NAMING I spent a lot of time considering the name for this module. It seems that a number of people across a number of languages all created similar functionality, though named very differently. For the benefit of keyword-based search tools and similar, here's a list of some other names this sort of object might be known by: * Extents * Overlays * Attribute or attributed strings * Markup * Out-of-band data CONSTRUCTOR new $st = String::Tagged->new( $str ); Returns a new instance of a String::Tagged object. It will contain no tags. If the optional $str argument is supplied, the string buffer will be initialised from this value. If $str is a String::Tagged object then it will be cloned, as if calling the clone method on it. new_tagged $st = String::Tagged->new_tagged( $str, %tags ); Shortcut for creating a new String::Tagged object with the given tags applied to the entire length. The tags will not be anchored at either end. clone (class) $new = String::Tagged->clone( $orig, %opts ); Returns a new instance of String::Tagged made by cloning the original, subject to the options provided. The returned instance will be in the requested class, which need not match the class of the original. The following options are recognised: only_tags => ARRAY If present, gives an ARRAY reference containing tag names. Only those tags named here will be copied; others will be ignored. except_tags => ARRAY If present, gives an ARRAY reference containing tag names. All tags will be copied except those named here. convert_tags => HASH If present, gives a HASH reference containing tag conversion functions. For any tags in the original to be copied whose names appear in the hash, the name and value are passed into the corresponding function, which should return an even-sized key/value list giving a tag, or a list of tags, to apply to the new clone. my @new_tags = $convert_tags->{$orig_name}->( $orig_name, $orig_value ); # Where @new_tags is ( $new_name, $new_value, $new_name_2, $new_value_2, ... ); As a further convenience, if the value for a given tag name is a plain string instead of a code reference, it gives the new name for the tag, and will be applied with its existing value. If only_tags is being used too, then the source names of any tags to be converted must also be listed there, or they will not be copied. start => INT Since version 0.22. Start at the given position; defaults to 0. end => INT Since version 0.22. End after the given position; defaults to end of string. This option overrides len. len => INT End after the given length beyond the start position; defaults to end of string. This option only applies if end is not given. clone (instance) $new = $orig->clone( %args ); Called as an instance (rather than a class) method, the newly-cloned instance is returned in the same class as the original. from_sprintf $str = String::Tagged->from_sprintf( $format, @args ); Since version 0.15. Returns a new instance of a String::Tagged object, initialised by formatting the supplied arguments using the supplied format. The $format string is similar to that supported by the core sprintf operator, though a few features such as out-of-order argument indexing and vector formatting are missing. This format string may be a plain perl string, or an instance of String::Tagged. In the latter case, any tags within it are preserved in the result. In the case of a %s conversion, the value of the argument consumed may itself be a String::Tagged instance. In this case it will be appended to the returned object, preserving any tags within it. All other conversions are handled individually by the core sprintf operator and appended to the result. join $str = String::Tagged->join( $sep, @parts ); Since version 0.17. Returns a new instance of a String::Tagged object, formed by concatenating each of the component piece together, joined with the separator string. The result will be much like the core join function, except that it will preserve tags in the resulting string. METHODS str $str = $st->str; $str = "$st"; Returns the plain string contained within the object. This method is also called for stringification; so the String::Tagged object can be used in a plain string interpolation such as my $message = String::Tagged->new( "Hello world" ); print "My message is $message\n"; length $len = $st->length; $len = length( $st ); Returns the length of the plain string. Because stringification works on this object class, the normal core length function works correctly on it. substr $str = $st->substr( $start, $len ); Returns a String::Tagged instance representing a section from within the given string, containing all the same tags at the same conceptual positions. plain_substr $str = $st->plain_substr( $start, $len ); Returns as a plain perl string, the substring at the given position. This will be the same string data as returned by substr, only as a plain string without the tags apply_tag $st->apply_tag( $start, $len, $name, $value ); Apply the named tag value to the given extent. The tag will start on the character at the $start index, and continue for the next $len characters. If $start is given as -1, the tag will be considered to start "before" the actual string. If $len is given as -1, the tag will be considered to end "after" end of the actual string. These special limits are used by set_substr when deciding whether to move a tag boundary. The start of any tag that starts "before" the string is never moved, even if more text is inserted at the beginning. Similarly, a tag which ends "after" the end of the string, will continue to the end even if more text is appended. This method returns the $st object. $st->apply_tag( $e, $name, $value ) Alternatively, an existing String::Tagged::Extent object can be passed as the first argument instead of two integers. The new tag will apply at the given extent. unapply_tag $st->unapply_tag( $start, $len, $name ); Unapply the named tag value from the given extent. If the tag extends beyond this extent, then any partial fragment of the tag will be left in the string. This method returns the $st object. $st->unapply_tag( $e, $name ); Alternatively, an existing String::Tagged::Extent object can be passed as the first argument instead of two integers. delete_tag $st->delete_tag( $start, $len, $name ); Delete the named tag within the given extent. Entire tags are removed, even if they extend beyond this extent. This method returns the $st object. $st->delete_tag( $e, $name ); Alternatively, an existing String::Tagged::Extent object can be passed as the first argument instead of two integers. delete_all_tag $st->delete_all_tag( $name ); Since version 0.21. Deletes every tag with the given name. This is more efficient than calling iter_extents to list the tags then delete_tag on each one individually in the case of a simple name match. This method returns the $st object. merge_tags $st->merge_tags( $eqsub ); Merge neighbouring or overlapping tags of the same name and equal values. For each pair of tags of the same name that apply on neighbouring or overlapping extents, the $eqsub callback is called, as $equal = $eqsub->( $name, $value_a, $value_b ); If this function returns true then the tags are merged. The equallity test function is free to perform any comparison of the values that may be relevant to the application; for example it may deeply compare referred structures and check for equivalence in some application-defined manner. In this case, the first tag of a pair is retained, the second is deleted. This may be relevant if the tag value is a reference to some object. iter_extents $st->iter_extents( $callback, %opts ); Iterate the tags stored in the string. For each tag, the CODE reference in $callback is invoked once, being passed a String::Tagged::Extent object that represents the extent of the tag. $callback->( $extent, $tagname, $tagvalue ); Options passed in %opts may include: start => INT Start at the given position; defaults to 0. end => INT End after the given position; defaults to end of string. This option overrides len. len => INT End after the given length beyond the start position; defaults to end of string. This option only applies if end is not given. only => ARRAY Select only the tags named in the given ARRAY reference. except => ARRAY Select all the tags except those named in the given ARRAY reference. Since version 0.21 it is safe to call delete_tag from within the callback function to remove the tag currently being iterated on. $str->iter_extents( sub { my ( $e, $n, $v ) = @_; $str->delete_tag( $e, $n ) if $n =~ m/^tmp_/; } ); Apart from this scenario, the tags in the string should not otherwise be added or removed while the iteration is occurring. iter_tags $st->iter_tags( $callback, %opts ); Iterate the tags stored in the string. For each tag, the CODE reference in $callback is invoked once, being passed the start point and length of the tag. $callback->( $start, $length, $tagname, $tagvalue ); Options passed in %opts are the same as for iter_extents. iter_extents_nooverlap $st->iter_extents_nooverlap( $callback, %opts ); Iterate non-overlapping extents of tags stored in the string. The CODE reference in $callback is invoked for each extent in the string where no tags change. The entire set of tags active in that extent is given to the callback. Because the extent covers possibly-multiple tags, it will not define the anchor_before and anchor_after flags. $callback->( $extent, %tags ); The callback will be invoked over the entire length of the string, including any extents with no tags applied. Options may be passed in %opts to control the range of the string iterated over, in the same way as the iter_extents method. If the only or except filters are applied, then only the tags that survive filtering will be present in the %tags hash. Tags that are excluded by the filtering will not be present, nor will their bounds be used to split the string into extents. iter_tags_nooverlap $st->iter_tags_nooverlap( $callback, %opts ); Iterate extents of the string using iter_extents_nooverlap, but passing the start and length of each extent to the callback instead of the extent object. $callback->( $start, $length, %tags ); Options may be passed in %opts to control the range of the string iterated over, in the same way as the iter_extents method. iter_substr_nooverlap $st->iter_substr_nooverlap( $callback, %opts ); Iterate extents of the string using iter_extents_nooverlap, but passing the substring of data instead of the extent object. $callback->( $substr, %tags ); Options may be passed in %opts to control the range of the string iterated over, in the same way as the iter_extents method. tagnames @names = $st->tagnames; Returns the set of tag names used in the string, in no particular order. get_tags_at $tags = $st->get_tags_at( $pos ); Returns a HASH reference of all the tag values active at the given position. get_tag_at $value = $st->get_tag_at( $pos, $name ); Returns the value of the named tag at the given position, or undef if the tag is not applied there. get_tag_extent $extent = $st->get_tag_extent( $pos, $name ); If the named tag applies to the given position, returns a String::Tagged::Extent object to represent the extent of the tag at that position. If it does not, undef is returned. If an extent is returned it will define the anchor_before and anchor_after flags if appropriate. get_tag_missing_extent $extent = $st->get_tag_missing_extent( $pos, $name ); If the named tag does not apply at the given position, returns the extent of the string around that position that does not have the tag. If it does exist, undef is returned. If an extent is returned it will not define the anchor_before and anchor_after flags, as these do not make sense for the range in which a tag is absent. set_substr $st->set_substr( $start, $len, $newstr ); Modifies a extent of the underlying plain string to that given. The extents of tags in the string are adjusted to cope with the modified region, and the adjustment in length. Tags entirely before the replaced extent remain unchanged. Tags entirely within the replaced extent are deleted. Tags entirely after the replaced extent are moved by appropriate amount to ensure they still apply to the same characters as before. Tags that start before and end after the extent remain, and have their lengths suitably adjusted. Tags that span just the start or end of the extent, but not both, are truncated, so as to remove the part of the tag applied on the modified extent but preserving that applied outside. If $newstr is a String::Tagged object, then its tags will be applied to $st as appropriate. Edge-anchored tags in $newstr will not be extended through $st, though they will apply as edge-anchored if they now sit at the edge of the new string. If $newstr is being appended to the end, then any existing edge-anchored tags at the end of $st are not extended through the string; they will instead become bounded to their end position before the append happened. If $newstr is otherwise treated as a plain string, then any existing edge-anchored tags at the end of $st are extended through the newly added content and will continue to be edge-anchored in the result. insert $st->insert( $start, $newstr ); Insert the given string at the given position. A shortcut around set_substr. If $newstr is a String::Tagged object, then its tags will be applied to $st as appropriate. If $start is 0, any before-anchored tags in will become before-anchored in $st. append $st->append( $newstr ); $st .= $newstr; Append to the underlying plain string. A shortcut around set_substr. If $newstr is a String::Tagged object, then its tags will be applied to $st as appropriate. Any after-anchored tags in will become after-anchored in $st. As per set_substr, whether any existing edge-anchored tags are extended through the newly-added content or become bounded to their current limit depends on whether $newstr is a String::Tagged instance or not. append_tagged $st->append_tagged( $newstr, %tags ); Append to the underlying plain string, and apply the given tags to the newly-inserted extent. Returns $st itself so that the method may be easily chained. concat $ret = $st->concat( $other ); $ret = $st . $other; Returns a new String::Tagged containing the two strings concatenated together, preserving any tags present. This method overloads normal string concatenation operator, so expressions involving String::Tagged values retain their tags. This method or operator tries to respect subclassing; preferring to return a new object of a subclass if either argument or operand is a subclass of String::Tagged. If they are both subclasses, it will prefer the type of the invocant or first operand. matches @subs = $st->matches( $regexp ); Returns a list of substrings (as String::Tagged instances) for every non-overlapping match of the given $regexp. This could be used, for example, to build a formatted string from a formatted template containing variable expansions: my $template = ... my %vars = ... my $ret = String::Tagged->new; foreach my $m ( $template->matches( qr/\$\w+|[^$]+/ ) ) { if( $m =~ m/^\$(\w+)$/ ) { $ret->append_tagged( $vars{$1}, %{ $m->get_tags_at( 0 ) } ); } else { $ret->append( $m ); } } This iterates segments of the template containing variables expansions starting with a $ symbol, and replaces them with values from the %vars hash, careful to preserve all the formatting tags from the original template string. match_extents @extents = $st->match_extents( $regexp ); Since version 0.20. Returns a list of extent objects for every non-overlapping match of the given $regexp. This is similar to "matches", except that the results are returned as extent objects instead of substrings, allowing access to the position information as well. If using the result of this method to find regions of a string to modify, remember that any length alterations will not update positions in later extent objects. However, since the extents are non-overlapping and in position order, this can be handled by iterating them in reverse order so that the modifications done first are later in the string. foreach my $e ( reverse $st->match_extents( $pattern ) ) { $st->set_substr( $e->start, $e->length, $replacement ); } split @parts = $st->split( $regexp, $limit ); Returns a list of substrings by applying the regexp to the string content; similar to the core perl split function. If $limit is supplied, the method will stop at that number of elements, returning the entire remainder of the input string as the final element. If the $regexp contains a capture group then the content of the first one will be added to the return list as well. sprintf $ret = $st->sprintf( @args ); Since version 0.15. Returns a new string by using the given instance as the format string for a "from_sprintf" constructor call. The returned instance will be of the same class as the invocant. debug_sprintf $ret = $st->debug_sprintf; Returns a representation of the string data and all the tags, suitable for debug printing or other similar use. This is a format such as is given in the DESCRIPTION section above. The output will consist of a number of lines, the first containing the plain underlying string, then one line per tag. The line shows the extent of the tag given by [---] markers, or a | in the special case of a tag covering only a single character. Special markings of < and > indicate tags which are "before" or "after" anchored. For example: Hello, world [---] word => 1 <[----------]> everywhere => 1 | space => 1 TODO * There are likely variations on the rules for set_substr that could equally apply to some uses of tagged strings. Consider whether the behaviour of modification is chosen per-method, per-tag, or per-string. * Consider how to implement a clone from one tag format to another which wants to merge multiple different source tags together into a single new one. AUTHOR Paul Evans <leonerd@leonerd.org.uk>