Do some calculation in a text file in shell - linux

Question

I have a text file:
$cat ifile.txt
this is a text file
assign x to 9 and y to 10.0702
define f(x)=x+y
I would like to disable the original line and divide the x-value by 2 and multiply the y-value by 2
My desired output is
$cat ofile.txt
this is a text file
#assign x to 9 and y to 10.0702
assign x to 5 and y to 20.1404
define f(x)=x+y
Here 5 is calculate from 9/2 and rounded to the next integer and 20.14 is calculated from 10.07x2 and not rounded
I am thinking of the following way, but can't write a script.
if [ line contains "assign x to" ]; then new_x_value=[next word]/2
if [ line contains "and y to" ]; then new_y_value=[next word]x2
if [ line contains "assign x to" ];
then disable it and add a line "assign x to new_x_value and y to new_y_value"

Answer 1

Would you please try the following:
#!/bin/bash
pat="(assign x to )([[:digit:]]+)( and y to )([[:digit:].]+)"
while IFS= read -r line; do
if [[ $line =~ $pat ]]; then
echo "#$line"
x2=$(echo "(${BASH_REMATCH[2]} + 1) / 2" | bc)
y2=$(echo "${BASH_REMATCH[4]} * 2" | bc)
echo "${BASH_REMATCH[1]}$x2${BASH_REMATCH[3]}$y2"
else
echo "$line"
fi
done < ifile.txt > ofile.txt
Output:
this is a text file
#assign x to 9 and y to 10.0702
assign x to 5 and y to 20.1404
define f(x)=x+y
The regex (assign x to )([[:digit:]]+)( and y to )([[:digit:].]+) matches
a literal string, followed by digits, followed by a literal string,
and followed by digits including decimal point.
The bc command (${BASH_REMATCH[2]} + 1) / 2 caclulates the ceiling
value of the input divided by 2.
The next bc command ${BASH_REMATCH[4]} * 2 multiplies the input by 2.
The reason I have picked bash is just because it supports back reference in regex and is easier to parse and reuse the input parameters than awk. As often pointed out, bash is not suitable for processing large files due to the performance reason. If you plan to large / multiple files, it will be recommended to use other languages like perl.
With perl you can say:
perl -pe 's|(assign x to )([0-9]+)( and y to )([0-9.]+)|
"#$&\n" . $1 . int(($2 + 1) / 2) . $3 . $4 * 2|ge' ifile.txt > ofile.txt
[EDIT]
If your ifile.txt looks like:
this is a text file
assign x to 9 and y to 10.0702 45
define f(x)=x+y
There are more than one space before the numbers.
One more value exists at the end (after whitespaces).
Then please try the following instead:
pat="(assign x to +)([[:digit:]]+)( and y to +)([[:digit:].]+)( +)([[:digit:].]+)"
while IFS= read -r line; do
if [[ $line =~ $pat ]]; then
echo "#$line"
x2=$(echo "(${BASH_REMATCH[2]} + 1) / 2" | bc)
y2=$(echo "${BASH_REMATCH[4]} * 2" | bc)
y3=$(echo "${BASH_REMATCH[6]} * 2" | bc)
echo "${BASH_REMATCH[1]}$x2${BASH_REMATCH[3]}$y2${BASH_REMATCH[5]}$y3"
else
echo "$line"
fi
done < ifile.txt > ofile.txt
Result:
this is a text file
#assign x to 9 and y to 10.0702 45
assign x to 5 and y to 20.1404 90
define f(x)=x+y
The plus sign after a whitespace is a regex quantifier and defines the number of repetition. In this case it matches one or more whitespace(s).

Answer 2

One in awk:
awk '
/assign/ { # when assign met in record
for(i=1;i<=NF-1;i++) # iterate from the beginning
if($i=="to" && $(i-1)=="x") # if to x
$(i+1)=((v=$(i+1)/2)>(u=int(v))?u+1:u) # ceil of division
else if($i=="to" && $(i-1)=="y") # if to y
$(i+1)*=2 # multiply by 2
}1' file # output
Output:
this is a text file
assign x to 5 and y to 20.1404
define f(x)=x+y
Sanity checking of the ceiling calculation left as homework...

Answer 3

awk '{if(match($0,/^assign/)){b=$0;split($0,a," ");a[8]=a[8]/2;a[4]=a[4]/2; for (x in a) {c = a[x] " " c; $0 = "#" b "\n" c } } { print }}'
Demo :
:>awk ' { if(match ($0, /^assign/)) {b=$0;split($0,a," ");a[8]=a[8]/2; a[4]=a[4]/2; for (x in a) {c = a[x] " " c; $0 = "#" b "\n" c } } { print }}' <ifile
this is a text file
#assign x to 9 and y to 10.0702
to x assign 5.0351 to y and 4.5
define f(x)=x+y
:>
Explanation:
awk ' {
if(match ($0, /^assign/)) <--- $0 is whole input record. ^ is start of line.
We are checking if record is starting with "assign"
{b=$0; <-- Assign input value to variable b
split($0,a," "); <-- Create a array by splitting input record with space as separator
a[8]=a[8]/2; a[4]=a[4]/2; <-- Divide value stored in 8 and 4 index
for (x in a) <-- Loop for getting all values of array
{c = a[x] " " c; <-- Create a variable by concatenating values of a
$0 = "#" b "\n" c <-- Update value of current record. "\n" new line operator
} }
{ print }}'