> > You mentioned earlier that you're expecting an integer, an integer, and > then a sequence of float. Don't count bytes if you can help it: let > Python's struct.calcsize() do this for you. > > http://www.python.org/doc/lib/module-struct.html > > Your machine may align bytes differently than you might expect: it may be > best to let the machine handle that for you.
Erk, Danny, is of course, right. Perhaps I could rewrite my earlier suggestion as - stepValPattern = ">2i" bytesToRead = struct.calcsize(stepValPattern) bytes = f.read(bytesToRead) (steps, value) = struct.unpack(stepValPattern, bytes) floatPattern = ">%df" % steps bytesToRead = struct.calcsize(floatPattern) bytes = f.read(bytesToRead) floatList = struct.unpack(floatPattern, bytes) f.read() works like following, (rough explanation follows) Let's say you have a 10 byte file - 00 0E A1 DC B3 09 FF AA B1 B2 00 f.read(3) will return 00 0E A1. Your file reference is now pointing at the 4th byte of the file, like so. 00 0E A1 DC B3 09 FF AA B1 B2 00 ...............^ <-current read position So calling f.read(3) again will return DC B3 09 Calling f.tell() will return the curernt read position, which would now be 6, the 7th byte (remembering computers count from 0 up.) f.seek(x) tells Python to move the read position to the xth byte of the file. You can use that with text files. I.E. If you've just called f.readlines(), you can use f.seek(0) to return to the start of the file. Anyway, I digress, but I'm trying to clarify how struct and f.read() work together. So, let's pretend all integers are 4 bytes, ditto all floats, and there's no byte alignment going on whatsoever. (of course, in real life you can't, hence using struct.calcsize()) So, you have a file like so - 03 00 00 00 0A 00 00 00 09 0B 21 CD.... That's step value step number of floats 03 00 00 00| 0A 00 00 00| 09 0B 21 CD.... ^ <-- read pos stepValPattern = ">2i" bytesToRead = struct.calcsize(stepValPattern) #bytesToRead is 8 (idealised ints) bytes = f.read(bytesToRead) At this point, bytes is 03 00 00 00 0A 00 00 00 (Python would show it as a string "\x03\x00\x00\x00\x0A\x00\x00\x00") And your file now looks like this - step value floats 03 00 00 00| 0A 00 00 00| 09 0B 21 CD.... .......................................^ <-- read pos Okay, so now to read the floats. (steps, value) = struct.unpack(stepValPattern, bytes) #Converts 03 00 00 00 0A 00 00 00 to (3, 9) 9 chosen at random #steps equals 3 So now we substitute steps into the floatPattern floatPattern = ">%df" % steps #floatPattern now equals ">3f" which is the same #struct pattern as ">fff" bytesToRead = struct.calcsize(floatPattern) #bytesToRead is 12 bytes = f.read(bytesToRead) floatList = struct.unpack(floatPattern, bytes) Does that make sense? I hope so... but suffice to say - struct.calcsize() ensures cross platform compatibility... I tend not to use the endian identifier unless I'm dealing with a data source that will always, absolutely be a certain endianess, saves having to rejig your patterns for each different platform. And use f.read(x) for binary data. Regards, Liam Clarke _______________________________________________ Tutor maillist - Tutor@python.org http://mail.python.org/mailman/listinfo/tutor
