Convert HTML/XML entities by default

lib/xml.ml

@@ -41,7 +41,273 @@ let _input_tree input : t =
   let data str : t = [`Data str] in
   input_tree ~el ~data input
 
-let of_string ?entity ?enc str =
+
+let entity_to_utf8_table =
+  let h = Hashtbl.create in
+  let open Hashtbl in
+  (* From https://en.wikipedia.org/wiki/List_of_XML_and_HTML_character_entity_references *)
+  add h "quot"   "\"";
+  add h "amp"    "&";
+  add h "apos"   "'";
+  add h "lt"     "<";
+  add h "gt"     ">";
+  add h "nbsp"   " ";
+  add h "iexcl"  "¡";
+  add h "cent"   "¢";
+  add h "pound"  "£";
+  add h "curren" "¤";
+  add h "yen"    "¥";
+  add h "brvbar" "¦";
+  add h "sect"   "§";
+  add h "uml"    "¨";
+  add h "copy"   "©";
+  add h "ordf"   "ª";
+  add h "laquo"  "«";
+  add h "not"    "¬";
+  add h "shy"    "­";
+  add h "reg"    "®";
+  add h "macr"   "¯";
+  add h "deg"    "°";
+  add h "plusmn" "±";
+  add h "sup2"   "²";
+  add h "sup3"   "³";
+  add h "acute"  "´";
+  add h "micro"  "µ";
+  add h "para"   "¶";
+  add h "middot" "·";
+  add h "cedil"  "¸";
+  add h "sup1"   "¹";
+  add h "ordm"   "º";
+  add h "raquo"  "»";
+  add h "frac14" "¼";
+  add h "frac12" "½";
+  add h "frac34" "¾";
+  add h "iquest" "¿";
+  add h "Agrave" "À";
+  add h "Aacute" "Á";
+  add h "Acirc"  "Â";
+  add h "Atilde" "Ã";
+  add h "Auml"   "Ä";
+  add h "Aring"  "Å";
+  add h "AElig"  "Æ";
+  add h "Ccedil" "Ç";
+  add h "Egrave" "È";
+  add h "Eacute" "É";
+  add h "Ecirc"  "Ê";
+  add h "Euml"   "Ë";
+  add h "Igrave" "Ì";
+  add h "Iacute" "Í";
+  add h "Icirc"  "Î";
+  add h "Iuml"   "Ï";
+  add h "ETH"    "Ð";
+  add h "Ntilde" "Ñ";
+  add h "Ograve" "Ò";
+  add h "Oacute" "Ó";
+  add h "Ocirc"  "Ô";
+  add h "Otilde" "Õ";
+  add h "Ouml"   "Ö";
+  add h "times"  "×";
+  add h "Oslash" "Ø";
+  add h "Ugrave" "Ù";
+  add h "Uacute" "Ú";
+  add h "Ucirc"  "Û";
+  add h "Uuml"   "Ü";
+  add h "Yacute" "Ý";
+  add h "THORN"  "Þ";
+  add h "szlig"  "ß";
+  add h "agrave" "à";
+  add h "aacute" "á";
+  add h "acirc"  "â";
+  add h "atilde" "ã";
+  add h "auml"   "ä";
+  add h "aring"  "å";
+  add h "aelig"  "æ";
+  add h "ccedil" "ç";
+  add h "egrave" "è";
+  add h "eacute" "é";
+  add h "ecirc"  "ê";
+  add h "euml"   "ë";
+  add h "igrave" "ì";
+  add h "iacute" "í";
+  add h "icirc"  "î";
+  add h "iuml"   "ï";
+  add h "eth"    "ð";
+  add h "ntilde" "ñ";
+  add h "ograve" "ò";
+  add h "oacute" "ó";
+  add h "ocirc"  "ô";
+  add h "otilde" "õ";
+  add h "ouml"   "ö";
+  add h "divide" "÷";
+  add h "oslash" "ø";
+  add h "ugrave" "ù";
+  add h "uacute" "ú";
+  add h "ucirc"  "û";
+  add h "uuml"   "ü";
+  add h "yacute" "ý";
+  add h "thorn"  "þ";
+  add h "yuml"   "ÿ";
+  add h "OElig"  "Œ";
+  add h "oelig"  "œ";
+  add h "Scaron" "Š";
+  add h "scaron" "š";
+  add h "Yuml"   "Ÿ";
+  add h "fnof"   "ƒ";
+  add h "circ"   "ˆ";
+  add h "tilde"  "˜";
+  add h "Alpha"  "Α";
+  add h "Beta"   "Β";
+  add h "Gamma"  "Γ";
+  add h "Delta"  "Δ";
+  add h "Epsilon" "Ε";
+  add h "Zeta"   "Ζ";
+  add h "Eta"    "Η";
+  add h "Theta"  "Θ";
+  add h "Iota"   "Ι";
+  add h "Kappa"  "Κ";
+  add h "Lambda" "Λ";
+  add h "Mu"     "Μ";
+  add h "Nu"     "Ν";
+  add h "Xi"     "Ξ";
+  add h "Omicron" "Ο";
+  add h "Pi"     "Π";
+  add h "Rho"    "Ρ";
+  add h "Sigma"  "Σ";
+  add h "Tau"    "Τ";
+  add h "Upsilon" "Υ";
+  add h "Phi"    "Φ";
+  add h "Chi"    "Χ";
+  add h "Psi"    "Ψ";
+  add h "Omega"  "Ω";
+  add h "alpha"  "α";
+  add h "beta"   "β";
+  add h "gamma"  "γ";
+  add h "delta"  "δ";
+  add h "epsilon" "ε";
+  add h "zeta"   "ζ";
+  add h "eta"    "η";
+  add h "theta"  "θ";
+  add h "iota"   "ι";
+  add h "kappa"  "κ";
+  add h "lambda" "λ";
+  add h "mu"     "μ";
+  add h "nu"     "ν";
+  add h "xi"     "ξ";
+  add h "omicron" "ο";
+  add h "pi"     "π";
+  add h "rho"    "ρ";
+  add h "sigmaf" "ς";
+  add h "sigma"  "σ";
+  add h "tau"    "τ";
+  add h "upsilon" "υ";
+  add h "phi"    "φ";
+  add h "chi"    "χ";
+  add h "psi"    "ψ";
+  add h "omega"  "ω";
+  add h "thetasym" "ϑ";
+  add h "upsih"  "ϒ";
+  add h "piv"    "ϖ";
+  add h "ensp"   " "; (* EN SPACE, U+2002 *)
+  add h "emsp"   " "; (* EM SPACE, U+2003 *)
+  add h "thinsp" " "; (* THIN SPACE, U+2009 *)
+  add h "zwnj"   "‌"; (* U+200C *)
+  add h "zwj"    "‍"; (* U+200D *)
+  add h "lrm"    "‎"; (* U+200E *)
+  add h "rlm"    "‏"; (* U+200F *)
+  add h "ndash"  "–";
+  add h "mdash"  "—";
+  add h "lsquo"  "‘";
+  add h "rsquo"  "’";
+  add h "sbquo"  "‚";
+  add h "ldquo"  "“";
+  add h "rdquo"  "”";
+  add h "bdquo"  "„";
+  add h "dagger" "†";
+  add h "Dagger" "‡";
+  add h "bull"   "•";
+  add h "hellip" "…";
+  add h "permil" "‰";
+  add h "prime"  "′";
+  add h "Prime"  "″";
+  add h "lsaquo" "‹";
+  add h "rsaquo" "›";
+  add h "oline"  "‾";
+  add h "frasl"  "⁄";
+  add h "euro"   "€";
+  add h "image"  "ℑ";
+  add h "weierp" "℘";
+  add h "real"   "ℜ";
+  add h "trade"  "™";
+  add h "alefsym" "ℵ";
+  add h "larr"   "←";
+  add h "uarr"   "↑";
+  add h "rarr"   "→";
+  add h "darr"   "↓";
+  add h "harr"   "↔";
+  add h "crarr"  "↵";
+  add h "lArr"   "⇐";
+  add h "uArr"   "⇑";
+  add h "rArr"   "⇒";
+  add h "dArr"   "⇓";
+  add h "hArr"   "⇔";
+  add h "forall" "∀";
+  add h "part"   "∂";
+  add h "exist"  "∃";
+  add h "empty"  "∅";
+  add h "nabla"  "∇";
+  add h "isin"   "∈";
+  add h "notin"  "∉";
+  add h "ni"     "∋"        ;
+  add h "prod"   "∏";
+  add h "sum"    "∑";
+  add h "minus"  "−";
+  add h "lowast" "∗";
+  add h "radic"  "√";
+  add h "prop"   "∝";
+  add h "infin"  "∞";
+  add h "ang"    "∠";
+  add h "and"    "∧";
+  add h "or"     "∨";
+  add h "cap"    "∩";
+  add h "cup"    "∪";
+  add h "int"    "∫";
+  add h "there4" "∴";
+  add h "sim"    "∼";
+  add h "cong"   "≅";
+  add h "asymp"  "≈";
+  add h "ne"     "≠";
+  add h "equiv"  "≡";
+  add h "le"     "≤";
+  add h "ge"     "≥";
+  add h "sub"    "⊂";
+  add h "sup"    "⊃";
+  add h "nsub"   "⊄";
+  add h "sube"   "⊆";
+  add h "supe"   "⊇";
+  add h "oplus"  "⊕";
+  add h "otimes" "⊗";
+  add h "perp"   "⊥";
+  add h "sdot"   "⋅";
+  add h "vellip" "⋮";
+  add h "lceil"  "⌈";
+  add h "rceil"  "⌉";
+  add h "lfloor" "⌊";
+  add h "rfloor" "⌋";
+  add h "lang"   "〈";
+  add h "rang"   "〉";
+  add h "loz"    "◊";
+  add h "spades" "♠";
+  add h "clubs"  "♣";
+  add h "hearts" "♥";
+  add h "diams"  "♦";
+  h
+
+let entity_to_utf8 entity =
+  try Some(Hashtbl.find entity_to_utf8_table entity)
+  with _ -> None
+
+
+let of_string ?(entity=entity_to_utf8) ?enc str =
   (* XXX: ugly hack to manually remove the DTD *)
   let remove_dtd str =
     let xml_decl = "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n" in

lib/xml.mli

@@ -26,3 +26,14 @@ val of_string :
   ?entity:(string -> string option) ->
   ?enc:encoding ->
   string -> t
+(** [of_string s] returns the XML tree described by [s].
+
+    @param entity is called to resolve non predefined entity
+    references such as "&".  It must return an UTF-8 string
+    corresponding to the replacement character data.  By default, the
+    {{:https://en.wikipedia.org/wiki/List_of_XML_and_HTML_character_entity_references}
+    standard entities} are recognized.
+
+    @param enc The encoding of the document.  Default [None] which
+    means that one does not know the encoding. *)
+;;