change a number in txt file - tcl

Question

I have the next expression: jj_ftfll h\\h\ -0.8898:0.006656 0.998:0.99999 h&j\hhh in a txt file,
and I need to add 0.005 to the 0.006656 number. I want to use Tcl and I can't think of any good idea.

Answer 1

There's several aspects that are tricky.
The file needs to be edited in-place despite the fact that the addition might change the length of the line. (Such an addition could potentially either lengthen or shorten the line.)
There needs to be a way of robustly recognising that that is the line to modify, and not some other line in the file. (This is actually the hardest of these problems in reality; it's extremely application-specific.)
The number needs to be extracted from the line, modified, and written back.
The values you are dealing with are potentially (well, actually) not represented precisely in IEEE binary floating point, which is what Tcl will use to do the calculations.
Bearing all that in mind, we are talking about these sorts of solutions:
We'll read the whole file in, split it into a list of strings, one string per line (henceforth referred to as the lines), update the lines of interest, and then write the whole lot back.
We'll use regexp to decide if a line is of interest. That's by far the most common command for this sort of task.
This one is messy in Tcl 8.6 and before. It's got a much better solution in Tcl 8.7.
There's really not all that much you can do about this. If you know the range of the numbers, you can use format to help… but it's messy. But maybe you'll get lucky.
set filename "foobar.txt"
# Get the lines of the file; this is GREAT if the file isn't too large
set f [open $filename]
set lines [split [read $f] "\n"]
close $f
# Now THAT'S what I call a horrible regular expression!
set RE {^(jj_ftfll\s+h\\\\h\\\s+-?[\d.]+:)(-?[\d.]+)(\s+-?[\d.]+:-?[\d.]+\s+h&j\hhh)$}
set newLines {}
foreach line $lines {
if {[regexp $RE $line -> prefix number suffix]} {
set line $prefix[expr {$number + 0.005}]$suffix
}
lappend newLines $line
}
# Write back over the file; the -nonewline prevents the number of lines from growing
set f [open $filename w]
puts -nonewline $f [join $newLines "\n"]
close $f
The trick with the regexp is that I am matching three pieces: the bit of the line before the part to replace (saved in the variable prefix), the number to replace itself (number), and the bit after the part to replace (suffix); the regexp command returns the number of times it matches (1 if the RE is found, 0 if it isn't). It's a scary RE mostly because it has -?[\d.]+ to match those floating point numbers, and I've changed spaces to \s+ (i.e., “at least one whitespace character”).
The version for 8.7 is this:
set filename "foobar.txt"
# Get the lines of the file; this is GREAT if the file isn't too large
set f [open $filename]
set lines [split [read $f] "\n"]
close $f
# Now THAT'S what I call a horrible regular expression!
set RE {^(jj_ftfll\s+h\\\\h\\\s+-?[\d.]+:)(-?[\d.]+)(\s+-?[\d.]+:-?[\d.]+\s+h&j\hhh)$}
proc addDeltaInLine {delta prefix number suffix} {
set number [expr {$number + $delta}]
return [string cat $prefix $number $suffix]
}
set newLines [lmap line $lines {
regsub -command $RE $line {addDeltaInLine 0.005}
}]
# Write back over the file; the -nonewline prevents the number of lines from growing
set f [open $filename w]
puts -nonewline $f [join $newLines "\n"]
close $f
The combination of lmap and regsub -command clean things up quite a bit. The RE is still scary though…