Although the section on word splitting in the Bash manual makes no distinction between leading and trailing null fields, Bash's word
splitting algorithm will not preserve a trailing null field. The reason for this appears to be the following logic in list_string
(much omitted):
for (result = (WORD_LIST *)NULL, sindex = 0; string[sindex]; )
{
.
.
current_word = string_extract_verbatim (string, slen, &sindex, separators);
.
.
ADVANCE_CHAR (string, slen, sindex);
.
.
}
If an IFS delimiter is the last char in string, string_extract_verbatim stops on it, ADVANCE_CHAR moves past it, and the for loop is
exited before string_extract_verbatim is called for the empty string between the final delimiter and the terminating Nul. The
following set of commands illustrates the problem:
$ IFS='|'
# Create a string that should split into 5 fields: leading null, 3 non-null
interior, and a trailing null
$ a='|a|b|c|'
$ set - $a
$ echo $#
4
$ echo "$1,$2,$3,$4"
,a,b,c
After reading the section on word splitting, I would have expected $# to equal 5 rather than 4, and I would have expected the result
of the final echo to be
,a,b,c,
But as you can see, there is a leading null field, but no trailing one. Is this
by design?
I guess the question is, what is meant by "delimits a field" in the following
excerpt from the Bash manual?
"Any character in IFS that is not IFS whitespace, along with any adjacent IFS
whitespace characters, delimits a field."
I suppose I'm interpreting it to mean "separates fields". It occurred to me that perhaps it means "terminates a field". The problem
with that definition, however, is that if I add any single character after the final '|' in the example above,
string_extract_verbatim will extract a final field, which is not terminated by anything in IFS, but simply by the end of the string.
In that case, the final IFS delimiter is separating the final two fields. The bottom line is that since the Bash manual does not
appear to distinguish between the cases of leading and trailing null fields, it appears that an arbitrary design choice determines
that leading null fields are kept, and trailing ones are not.
Thanks,
Brett S.
_______________________________________________
Bug-bash mailing list
Bug-bash@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-bash
