>Which is, in the end, a lot more work than the low-level IO functions. Why plow through the file twice when you can do it once and disk access is one of the slowest things that we do in VFP.
True, but most of the time these are some one-time conversions, and using an interim table is a good way to take a look at the converted data. I usually change the structure of this table to keep some temporary data just to see the way they are laid out (like Cobol six-digit dates - they may be dmy, mdy, ymd), and to check whether they really correspond to the documented structure (which they usually don't do - there's always some undocumented change :)
Even when I have to use LLFF to read the data and chop them into memvars to insert, I Create Table where I insert them into, have the table at least visually checked, and then append into the real table. Saves some time while trying to guess the original table design.