[#CLJ-945] clojure.string/capitalize can give wrong result if first char is supplementary - Clojure JIRA

Details

Type: Defect
Status: Closed
Priority: Minor
Resolution: Declined
Affects Version/s: Release 1.2, Release 1.3, Release 1.4
Fix Version/s: Release 1.5
Component/s: None
Labels:
None
Environment:
all

Patch:
Code and Test
Approval:
Not Approved

Description

When the first unicode code point of a string is supplementary (i.e. requires two 16-bit Java chars to represent in UTF-16), and that first code point is changed by converting it to upper case, clojure.string/capitalize gives the wrong answer.

More Options
- Sort By Name
- Sort By Date
- Ascending
- Descending
- Download All

Attachments

capitalize-for-supplementary-chars-patch.txt

05/Mar/12 11:33 AM

2 kB

Andy Fingerhut

Activity

Ascending order - Click to sort in descending order

Rich Hickey made changes - 13/Apr/12 9:18 AM

Field	Original Value	New Value
Fix Version/s		Release 1.5 [ 10150 ]

Stuart Halloway made changes - 08/Jun/12 10:15 AM

Approval

Screened [ 10004 ]

Hide

Permalink

Rich Hickey added a comment - 20/Jul/12 7:43 AM

Isn't this a Java bug?

Show

Rich Hickey added a comment - 20/Jul/12 7:43 AM Isn't this a Java bug?

Rich Hickey made changes - 20/Jul/12 7:43 AM

Approval

Screened [ 10004 ]

None [ 10008 ]

Hide

Permalink

Andy Fingerhut added a comment - 20/Jul/12 12:36 PM

If using UTF-16 to encode Unicode strings, and making every UTF-16 code unit (i.e. Java char) individually indexable as a separate entity in strings, is such a bad design choice that you consider it a bug, then yes, this is a Java bug (and a bug in all the other systems that use UTF-16 in this way).

clojure.string/capitalize isn't using some Java capitalization method that has a bug, though. By calling (.toUpperCase (subs s 0 1)) it is not giving enough information to .toUpperCase for any implementation, Java or otherwise, to do the job correctly. It is analogous to calling toupper on the least significant 4 bits of the ASCII encoding of a letter and expecting it to return the correct answer.

Show

Andy Fingerhut added a comment - 20/Jul/12 12:36 PM If using UTF-16 to encode Unicode strings, and making every UTF-16 code unit (i.e. Java char) individually indexable as a separate entity in strings, is such a bad design choice that you consider it a bug, then yes, this is a Java bug (and a bug in all the other systems that use UTF-16 in this way). clojure.string/capitalize isn't using some Java capitalization method that has a bug, though. By calling (.toUpperCase (subs s 0 1)) it is not giving enough information to .toUpperCase for any implementation, Java or otherwise, to do the job correctly. It is analogous to calling toupper on the least significant 4 bits of the ASCII encoding of a letter and expecting it to return the correct answer.

Stuart Halloway made changes - 01/Mar/13 9:43 AM

Resolution		Declined [ 2 ]
Status	Open [ 1 ]	Closed [ 6 ]

People

Assignee:

Unassigned

Reporter:

Andy Fingerhut

Vote (0)

Watch (1)

Dates

Created:

05/Mar/12 11:33 AM

Updated:

01/Mar/13 9:43 AM

Resolved:

01/Mar/13 9:43 AM