diff options
Diffstat (limited to 'java/src/com/android/inputmethod/latin/StringUtils.java')
| -rw-r--r-- | java/src/com/android/inputmethod/latin/StringUtils.java | 140 |
1 files changed, 137 insertions, 3 deletions
diff --git a/java/src/com/android/inputmethod/latin/StringUtils.java b/java/src/com/android/inputmethod/latin/StringUtils.java index 59ad28fc9..d5ee58a63 100644 --- a/java/src/com/android/inputmethod/latin/StringUtils.java +++ b/java/src/com/android/inputmethod/latin/StringUtils.java @@ -106,10 +106,19 @@ public final class StringUtils { } } - public static String toTitleCase(final String s, final Locale locale) { + public static String capitalizeFirstCodePoint(final String s, final Locale locale) { if (s.length() <= 1) { - // TODO: is this really correct? Shouldn't this be s.toUpperCase()? - return s; + return s.toUpperCase(locale); + } + // Please refer to the comment below in + // {@link #capitalizeFirstAndDowncaseRest(String,Locale)} as this has the same shortcomings + final int cutoff = s.offsetByCodePoints(0, 1); + return s.substring(0, cutoff).toUpperCase(locale) + s.substring(cutoff); + } + + public static String capitalizeFirstAndDowncaseRest(final String s, final Locale locale) { + if (s.length() <= 1) { + return s.toUpperCase(locale); } // TODO: fix the bugs below // - This does not work for Greek, because it returns upper case instead of title case. @@ -213,4 +222,129 @@ public final class StringUtils { if (1 == capsCount) return CAPITALIZE_FIRST; return (letterCount == capsCount ? CAPITALIZE_ALL : CAPITALIZE_NONE); } + + public static boolean isIdenticalAfterUpcase(final String text) { + final int len = text.length(); + for (int i = 0; i < len; i = text.offsetByCodePoints(i, 1)) { + final int codePoint = text.codePointAt(i); + if (Character.isLetter(codePoint) && !Character.isUpperCase(codePoint)) { + return false; + } + } + return true; + } + + public static boolean isIdenticalAfterDowncase(final String text) { + final int len = text.length(); + for (int i = 0; i < len; i = text.offsetByCodePoints(i, 1)) { + final int codePoint = text.codePointAt(i); + if (Character.isLetter(codePoint) && !Character.isLowerCase(codePoint)) { + return false; + } + } + return true; + } + + public static boolean isIdenticalAfterCapitalizeEachWord(final String text, + final String separators) { + boolean needCapsNext = true; + final int len = text.length(); + for (int i = 0; i < len; i = text.offsetByCodePoints(i, 1)) { + final int codePoint = text.codePointAt(i); + if (Character.isLetter(codePoint)) { + if ((needCapsNext && !Character.isUpperCase(codePoint)) + || (!needCapsNext && !Character.isLowerCase(codePoint))) { + return false; + } + } + // We need a capital letter next if this is a separator. + needCapsNext = (-1 != separators.indexOf(codePoint)); + } + return true; + } + + // TODO: like capitalizeFirst*, this does not work perfectly for Dutch because of the IJ digraph + // which should be capitalized together in *some* cases. + public static String capitalizeEachWord(final String text, final String separators, + final Locale locale) { + final StringBuilder builder = new StringBuilder(); + boolean needCapsNext = true; + final int len = text.length(); + for (int i = 0; i < len; i = text.offsetByCodePoints(i, 1)) { + final String nextChar = text.substring(i, text.offsetByCodePoints(i, 1)); + if (needCapsNext) { + builder.append(nextChar.toUpperCase(locale)); + } else { + builder.append(nextChar.toLowerCase(locale)); + } + // We need a capital letter next if this is a separator. + needCapsNext = (-1 != separators.indexOf(nextChar.codePointAt(0))); + } + return builder.toString(); + } + + /** + * Approximates whether the text before the cursor looks like a URL. + * + * This is not foolproof, but it should work well in the practice. + * Essentially it walks backward from the cursor until it finds something that's not a letter, + * digit, or common URL symbol like underscore. If it hasn't found a period yet, then it + * does not look like a URL. + * If the text: + * - starts with www and contains a period + * - starts with a slash preceded by either a slash, whitespace, or start-of-string + * Then it looks like a URL and we return true. Otherwise, we return false. + * + * Note: this method is called quite often, and should be fast. + * + * TODO: This will return that "abc./def" and ".abc/def" look like URLs to keep down the + * code complexity, but ideally it should not. It's acceptable for now. + */ + public static boolean lastPartLooksLikeURL(final CharSequence text) { + int i = text.length(); + if (0 == i) return false; + int wCount = 0; + int slashCount = 0; + boolean hasSlash = false; + boolean hasPeriod = false; + int codePoint = 0; + while (i > 0) { + codePoint = Character.codePointBefore(text, i); + if (codePoint < Constants.CODE_PERIOD || codePoint > 'z') { + // Handwavy heuristic to see if that's a URL character. Anything between period + // and z. This includes all lower- and upper-case ascii letters, period, + // underscore, arrobase, question mark, equal sign. It excludes spaces, exclamation + // marks, double quotes... + // Anything that's not a URL-like character causes us to break from here and + // evaluate normally. + break; + } + if (Constants.CODE_PERIOD == codePoint) { + hasPeriod = true; + } + if (Constants.CODE_SLASH == codePoint) { + hasSlash = true; + if (2 == ++slashCount) { + return true; + } + } else { + slashCount = 0; + } + if ('w' == codePoint) { + ++wCount; + } else { + wCount = 0; + } + i = Character.offsetByCodePoints(text, i, -1); + } + // End of the text run. + // If it starts with www and includes a period, then it looks like a URL. + if (wCount >= 3 && hasPeriod) return true; + // If it starts with a slash, and the code point before is whitespace, it looks like an URL. + if (1 == slashCount && (0 == i || Character.isWhitespace(codePoint))) return true; + // If it has both a period and a slash, it looks like an URL. + if (hasPeriod && hasSlash) return true; + // Otherwise, it doesn't look like an URL. + return false; + } } |