| # Changes |
| |
| The Unicode Utilities have been modified to support both properties from the |
| released version of Unicode (via ICU) and from the new Unicode beta. |
| |
| To get the beta version of the property, insert β *after* the property name. |
| Examples: |
| |
| | `\p{Word_Break=ALetter}` | Released version of Unicode | |
| | `\p{Word_Breakβ=ALetter}` | Beta version of Unicode | |
| |
| |
| For example, to see additions to that property value in the beta version, use: |
| |
| <center> |
| |
| [`\p{Word_Breakβ=ALetter}-\\p{Word_Break=ALetter}`](https://util.unicode.org/UnicodeJsps/list-unicodeset.jsp?a=%5Cp%7BWord_Break%CE%B2%3DALetter%7D-%5Cp%7BWord_Break%3DALetter%7D&g=&i=) |
| |
| </center> |
| |
| |
| ## Caveats |
| |
| The support is not complete done, and there are some known problems. |
| |
| 1. Some properties are not supported in beta versions. See |
| <https://util.unicode.org/UnicodeJsps/properties.jsp> |
| for the list. |
| 2. When characters are listed, the new blocks and subheads don't show up. |
| 3. If you use a property that has a β version but no ICU version, you get no |
| error: just an empty listing. |
| 4. The beta properties don't yet have the "shorthands" for cases like \\p{Lu}. |
| So make sure the property is listed, eg \\p{gcβ=Lu} |
| 1. Example: |
| [`\p{gcβ=Lu}-\\p{gc=Lu}`](https://util.unicode.org/UnicodeJsps/list-unicodeset.jsp?a=%5Cp%7Bgc%CE%B2%3DLu%7D-%5Cp%7Bgc%3DLu%7D&g=&i=) |
| 5. Tools for segmentation, etc. use the release properties; there isn't a way |
| to have them use the beta properties. |
| 6. There are probably others... |
| |
| If you find a problem, please file a ticket at |
| <https://cldr.unicode.org/index/bug-reports>: make sure to start the summary with |
| "Unicode Utilities: " |
| |
| [Back to Unicode Utilities Help Home](index) |