From: Logan Shaw [EMAIL PROTECTED]
Chrix Bell wrote:
I have make a struct with two offsets and a take-place field
looks like this:
struct whatever{
unsigned Aoffset;
unsigned Boffset;
char FirstField;
}
I want to use this to pack the record size,
What does pack the record size mean?
Since the original message described writing two fields my guess
would be to pack string data from multiple fields into a database
record without unnecessary padding. This is hardly a new concept
for Palm OS databases :-)
I tend to create the struct ending with a char[2] to indicate that this
is string data rather than an individual char and, more importantly,
comment the struct to indicate to any programmer that follows you
that the record data entends past the end of the struct.
data would be write from
FirstField, write A, then from (FirstField+ Boffset), write B.
What are A and B? Are the some sort of data you want to write
into a database? Are they variables? Are they instances of
struct whatever?
Doesn't matter. The concepts are the same regardless of the
objects - find the size of the first and second, create a buffer
big enough for the data and any management overhead, write
the first record to the buffer at offset zero (or just past any record
header), find the size of the second and write at an offset just
past the end of the first block. Then (if you want to use the technique
this poster seems to be following) write the offset for the second
record to the header at the start of the buffer and close the record.
The problem is that you don't write B at (FirstField+ Boffset).
Generally each field is written at the stored offset matching it
so A would be stored at Aoffset (strictly redundant because
you know it starts at the offset of the FirstField struct member),
then Boffset is calculated as Aoffset + (sizeof(A) or StrLen(A))
+ 1 (depending on whether you're storing a struct or a string and
a nasty trap for beginnners) and B is stored at Boffset. Continue
as required. The calculation mistake is probably why you're
getting an out of bound error.
This has been done before so try to find some code that
implements this and copy it.
I am wondering if there are some other kind of skills can make
the record's length not a fixed one.
If by the record you mean the struct, then it is impossible.
In C, a struct's size is determined at compile time.
Strictly correct, but slightly misleading...
If you need
to write a data structure whose length differs depending on what it
contains, you will need to write it by hand, one byte at a time.
(Or with DmWrite() or MemMove() or similar.)
I've seen this technique of a using a struct to describe the fixed
elements and appending strings to the end of it fairly often in
Palm OS apps - I'm sure at least one of the sample apps does
things this way so it's not only not impossible, it's fairly standard
for Palm OS. The confusion is that you've got to understand
that the struct is just a guide that simplifies accessing the start
of the record - the actual data extends beyond it so, as you say,
you've got to be careful about deciding whether you're going to
write the struct as a block and then append data or write the
contents of the struct member by member (generally safer).
If you're only storing strings you actually don't need the offsets
as long as you make sure that each string is null-terminated in the buffer.
You can then read the first string at offset zero and the second
at offset StrLen(firststring) + 1).
Storing the offsets can still be useful if you're writing many strings
and want to quickly extract one without reading the preceeding
strings.
Chris Tutty
--
For information on using the PalmSource Developer Forums, or to unsubscribe,
please see http://www.palmos.com/dev/support/forums/
