Module: AcroForge::Constants

Defined in:

lib/acroforge/constants.rb

Constant Summary

TYPO_PHRASE_REPLACEMENTS =

{
  "identi_fi_cation" => "identification",
  "identi_cation" => "identification",
  "ide_ntity" => "identity",
  "contribu_on" => "contribution",
  "con_rmed" => "confirmed",
  "na_onal" => "national",
  "ocial" => "official",
  "modeof" => "mode_of",
  "modeofr" => "mode_of_r",
  "nameof" => "name_of"
}.freeze

UNICODE_REPLACEMENTS =

PDF text extraction returns Unicode quirks: ligatures (fi instead of f+i), fullwidth letters, curly quotes, etc. AllTextProcessor normalizes via Unicode NFKC first, which handles most of the “compatibility” subset automatically (ligatures, fullwidth, superscript digits, …). NFKC does NOT touch these characters — they’re separate codepoints, not compatibility decompositions — so we substitute them manually.

{
  "\u{2018}" => "'",      # left single quote
  "\u{2019}" => "'",      # right single quote
  "\u{201A}" => "'",      # single low-9 quote
  "\u{201C}" => '"',      # left double quote
  "\u{201D}" => '"',      # right double quote
  "\u{201E}" => '"',      # double low-9 quote
  "\u{2013}" => "-",      # en dash
  "\u{2014}" => "-",      # em dash
  "\u{2010}" => "-",      # hyphen
  "\u{2011}" => "-",      # non-breaking hyphen
  "\u{2212}" => "-",      # minus sign
  "\u{00AD}" => "",       # soft hyphen (often invisible artifact)
  "\u{200B}" => "",       # zero-width space
  "\u{200C}" => "",       # zero-width non-joiner
  "\u{200D}" => "",       # zero-width joiner
  "\u{FEFF}" => "",       # zero-width no-break space (BOM)
  "\u{2026}" => "...",    # ellipsis
  "\u{2022}" => "*",      # bullet
  "\u{00B7}" => "*"       # middle dot used as bullet
}.freeze