"Amy L." <am**@paxemail.com> wrote in message
news:ua**************@TK2MSFTNGP10.phx.gbl...
Is there a way through .net to read a very large text file (400MB+)
backwards line by line. In system.io the filestream class has a "seek"
method but the only read method requires you to know how many bytes to
read in.
My problem is that the line length of this log file is not constant so
there is no easy way to read one line in. The only thing that is constant is
that each line is terminated by a carriage return.
If my only option was to use the filestream class I suppose I can read
data back in chunks and try to parse the data and determine lines...
Is there a better way?
Amy.
About all you can do is seek to, say, 1024 bytes before the end; read 1024
bytes; work through them byte by byte to chop up into lines; seek to 1024
bytes before that; read 1024 bytes; and keep going.
The problem is that, as you know, "lines" mean nothing to the low-level disk
functions in the operating system. This isn't like an IBM mainframe where
files have a "record length" and are divided into "card images".
I suggested 1024 because the cluster size (allocation unit) is probably a
multiple of 1024, so you will be synchronized with the disk sector
boundaries; that should be mildly advantageous as regards speed.