#30
  • Created on
    2009-09-10 12:36
  • Reported by
    André Thieme
  • Status
    Invalid
  • Priority
    Normal (3)
  • Assigned to
    stuart.halloway
  • Milestone
    Release 1.2
  • Plan Level
    -
  • Permission type
    Private
  • Due Date
    -

Attachments

Related Tickets

Add new subtask or bug

append-spit should only write out an encoding marker once

In clojure.contrib.duck-streams append-spit writes out encoding
markers (for UnicodeLittle for example this is a FEFF in hex)
each time it appends to a file. This should happen only when
the file is initially created.

Test case for reproducing this behaviour:

(use 'clojure.contrib.duck-streams)

(binding [*default-encoding* "UnicodeLittle"]
  (append-spit "/foo.txt" "Line 1\n"))
(binding [*default-encoding* "UnicodeLittle"]
  (append-spit "/foo.txt" "Line 2\n"))

(slurp "c:/foo.txt" "UnicodeLittle")


The slurp outputs
"Line 1\n?Line 2\n"
The expected output is:
"Line 1\nLine 2\n"
User picture
  on 2010-01-31 03:33 *
By stuartsierra
Milestone set to Backlog
Updating tickets (#1, #2, #3, #4, #6, #20, #23, #25, #30, #31, #33, #34, #35, #37, #38, #52, #55, #58, #59, #60, #61, #62, #63, #64)
User picture
  on 2010-04-13 12:18 *
By stuart.halloway
Assigned to set to stuart.halloway
Milestone changed from Backlog to Release 1.2
User picture
  on 2010-04-13 17:08 *
By stuart.halloway
Status changed from New to Invalid
I am not sure there is a good answer here. The code above chooses an encoding with an explicit marker, and gets what it asks for. :-(

One proposed solution (http://github.com/sergey-miryanov/clojure-contrib/commits/bug-30) tries to detect this scenario, and recover via a hard-coded mapping between encodings-with-markers and similar-encodings-without. But I don't think this can work in general, because the set of possible encodings is open and the Charset API doesn't provide a mapping between the with-markers and without-markers versions.

Sorry, and please feel free to reopen this if I am missing an obvious approach.