Comparison Profile

Filters and optimizations

Filter content

The Filter content GUI page describes the current profile sidebar. This Programming page documents raw values and Java regex execution separately from the user interface.

Raw profile values

Name	Value and meaning
`FILTERS`	Contains the `REGEXP` filter identifier to activate the content filter. Values are stored as a comma-separated list.
`FILTER_PATTERNS`	Contains multiple pattern lines separated by literal line breaks. The canonical format of each line is `pattern\|type\|status`, with `type` as `text` or `regexp` and `status` as `active` or `inactive`.
`REGEX_MATCH_WORDS`	Boolean value, default `true`. With `true`, every word touched by a match is removed completely. With `false`, only the matching character ranges are removed.

Inactive patterns remain stored but are not applied. Active patterns must contain content. An active regex pattern must compile as a Java pattern; an invalid inactive pattern can remain as a draft.

Line breaks and profile XML

When a pattern is saved, literal line breaks inside the pattern are encoded as CR. The decoder also understands the legacy token BR. If a pattern itself contains one of these tokens or starts with the reserved prefix, the codec uses the versioned envelope <base64> so that the pattern text remains unambiguous.

<entry key="FILTERS">REGEXP</entry>
<entry key="REGEX_MATCH_WORDS">true</entry>
<entry key="FILTER_PATTERNS">Version [[CR]][[CR]]|text|active
(?s)Start.*?End|regexp|active</entry>

The lines in FILTER_PATTERNS are processed in their stored order. Type text is matched literally; type regexp is compiled as a Java regex.

Structure and API keys

Key	Mapping
`pdfc.filter.regex_filter`	Adds or removes `REGEXP` in `FILTERS`.
`pdfc.filter.regex_filter.filter_patterns`	Accesses `FILTER_PATTERNS` as a list of pattern lines.
`pdfc.filter.regex_filter.match_words`	Accesses `REGEX_MATCH_WORDS`. `true` corresponds to Whole words; `false` corresponds to Partial matches.

Java regex semantics

Regex patterns are compiled with Java Pattern.compile without a fixed flag mask. Plain-text patterns are quoted with Java Pattern.quote before compilation. Java regular-expression rules therefore apply, including embedded flags.

(?s) enables DOTALL for the following expression and lets the dot match across line breaks.
(?m) enables MULTILINE and changes the meaning of ^ and $ for individual lines.
Other Java flags can be embedded, for example (?i) for a case-insensitive search.

(?s)Start.*?End
(?m)^Article:.*$

Each active pattern is applied with Java Matcher.find() to the complete extracted page text. Matching is not performed separately for individual PDF text elements. An invalid active Java regex is rejected during sidebar validation.

Capture-group semantics

If a match contains no capture groups, the complete match, group 0, is removed. If the pattern contains at least one capture group, only participating groups 1 through n are removed. Group 0 is not removed in addition. An optional group that does not participate in a particular match is skipped.

Invoice (number: [0-9]+)

With Whole words, every word touched by a match is removed completely. With Partial matches, only the character ranges described by the match or participating capture groups are removed. Overlapping ranges are merged before page elements are adjusted.

Filter content controls the REGEXP filter identifier in FILTERS.
Filter Patterns is an editable list. Add filter ... creates an active pattern as Plain text or Regular Expression.
Enable filter controls the active or inactive status of each pattern row.
Type with Plain text uses text and matches the content literally. Regular Expression uses regexp and Java regex semantics.
Pattern writes the pattern portion. Exclude whole words with Whole words or Partial matches writes REGEX_MATCH_WORDS.
When an empty pattern is activated or an invalid regular expression is entered, the sidebar displays a validation message for the affected row.