ben:
thanks for the feedback. SPSS format documentation is pretty sparse
(only good paper I found is at
http://www.wotsit.org/list.asp?search=sav&button=GO!) and I have this in
my notes (after browsing the web for a while). Also, the sfm-read.c
file I have in the pspp source code seem to expect a layout_code 2
(around line 650). My source code may be outdated though.
I'm using this check in a Java SPSS Reader but will then just change the
code to accept both 2 and 3. I likewise found no difference in the test
files I have received.
best
Pascal
/* Check endianness. */
if (hdr.layout_code == 2)
r->reverse_endian = 0;
else
{
bswap_int32 (&hdr.layout_code);
if (hdr.layout_code != 2)
lose ((ME, _("%s: File layout code has unexpected value %d. Value "
"should be 2, in big-endian or little-endian format."),
handle_get_filename (r->fh), hdr.layout_code));
Ben Pfaff wrote:
Pascal Heus <[EMAIL PROTECTED]> writes:
Does anyone now the exact meaning of the record layout type byte
present in the SPSS .sav file header? This is byte #65 that follows
the $FL2 signature and the 60-byte information string. According to
the documentation I found and based on source codes, the only allowed
values are 2=SPSS file and 3=portable file. Problem is that I have
recently received a few files produced I think with SPSS 12 or 13 that
have a layout code = 3 (not sure why) but look exactly like a record
layout 2 file. I'm trying to figure out where this difference comes
What source have you found that says 3 = portable file? I do not
recall reading that anywhere.
I have a fairly extensive library of .sav files that I have found
around the net. A few of them have record layout 3; most of them
have record layout 2. There does not seem to be any difference
in the format of the rest of the file based on the record layout
byte.
PSPP currently just uses the presence of a value 2 or 3 in this
field to detect whether the file is little-endian or big-endian
and otherwise ignores it.
_______________________________________________
pspp-dev mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/pspp-dev
