- for (String token : tokens) {
- token = TRIM_PUNC.matcher(token).replaceAll("");
- if (/*!alreadyDone.contains(token) && */token.length() > 0) {
- indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), entryTypeName);
- // alreadyDone.add(token);
-
- // also split words on dashes, do them, too.
- if (token.contains("-")) {
- final String[] dashed = token.split("-");
- for (final String dashedToken : dashed) {
- if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
- indexBuilder.addEntryWithTokens(entryData, Collections.singleton(dashedToken), EntryTypeName.PART_OF_HYPHENATED);
+ private StringBuilder extractParenthesized(StringBuilder in, String startChar, String endChar) {
+ StringBuilder res = new StringBuilder();
+ int pos = 0;
+ while ((pos = in.indexOf(startChar, pos)) != -1) {
+ int end = in.indexOf(endChar, pos + 1);
+ if (end == -1) break;
+ res.append(in, pos + 1, end).append(" ");
+ in.replace(pos, end + 1, " ");
+ pos++; // skip the just appended space
+ }
+ return res;
+ }
+
+ private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
+ final IndexedEntry entryData, final int subfieldIdx, final int numSubFields) {
+ final StringBuilder fieldsb = new StringBuilder(field);
+ // remove bracketed and parenthesized stuff.
+ final StringBuilder bracketed = extractParenthesized(fieldsb, "[", "]");
+ final StringBuilder parenthesized = extractParenthesized(fieldsb, "(", ")");
+
+ field = fieldsb.toString().trim();
+
+ // split words on non -A-z0-9, do them.
+ final String[] tokens = NON_CHAR_DASH.split(field);
+
+ final EntryTypeName entryTypeName;
+ if (numSubFields == 1) {
+ assert subfieldIdx == 0;
+ if (tokens.length == 1) {
+ entryTypeName = EntryTypeName.ONE_WORD;
+ } else if (tokens.length == 2) {
+ entryTypeName = EntryTypeName.TWO_WORDS;
+ } else if (tokens.length == 3) {
+ entryTypeName = EntryTypeName.THREE_WORDS;
+ } else if (tokens.length == 4) {
+ entryTypeName = EntryTypeName.FOUR_WORDS;
+ } else {
+ entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
+ }
+ } else {
+ assert numSubFields > 1;
+ if (subfieldIdx == 0) {
+ if (tokens.length == 1) {
+ entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
+ } else {
+ entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
+ }
+ } else {
+ assert subfieldIdx > 0;
+ if (tokens.length == 1) {
+ entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
+ } else {
+ entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
+ }
+ }
+ }
+
+ for (String token : tokens) {
+ token = TRIM_PUNC.matcher(token).replaceAll("");
+ if (/*!alreadyDone.contains(token) && */!token.isEmpty()) {
+ indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), entryTypeName);
+ // alreadyDone.add(token);
+
+ // also split words on dashes, do them, too.
+ if (token.indexOf('-') != -1) {
+ final String[] dashed = StringUtil.split(token, "-");
+ for (final String dashedToken : dashed) {
+ if (/*!alreadyDone.contains(dashedToken) && */!dashedToken.isEmpty()) {
+ indexBuilder.addEntryWithTokens(entryData, Collections.singleton(dashedToken), EntryTypeName.PART_OF_HYPHENATED);
+ }
+ }
+ }
+
+ } // if (!alreadyDone.contains(token)) {
+ } // for (final String token : tokens) {
+
+ // process bracketed stuff (split on spaces and dashes always)
+ if (bracketed.length() > 0) {
+ final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
+ for (final String token : bracketedTokens) {
+ assert token.indexOf("-") == -1;
+ if (/*!alreadyDone.contains(token) && */!token.isEmpty()) {
+ indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), EntryTypeName.BRACKETED);
+ }
+ }
+ }
+
+ // process paren stuff
+ if (parenthesized.length() > 0) {
+ final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
+ for (final String token : parenTokens) {
+ assert token.indexOf("-") == -1;
+ if (/*!alreadyDone.contains(token) && */!token.isEmpty()) {
+ indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), EntryTypeName.PARENTHESIZED);
+ }