浮点比较问题

Question

I am trying to check if a value I read from a text file is zero:

[[ $(echo $line | cut -d" " -f5) -gt 0 ]] && [[ $(echo $line | cut -d" " -f7 | bc -l) -eq 0 ]]

With the first condition there is no problem because f5 are integers. The problem comes form the second condition. I receive this error message:

[[: 1.235: syntax error: invalid arithmetic operator (error token is ".235")

I have tried several suggestions I found in different forums such as using echo $line | cut -d" " -f7 | bc -l echo $line | cut -d" " -f7 | bc -l echo $line | cut -d" " -f7 | bc -l with and without double quotes, etc. However, the error persist. f7 is a positive number and is given with 3 decimal places. Removing decimals or approximating is not an option because I need the result to be exactly zero (0.000).

Answer 1

Generally, you can't compare floating-point numbers for equality. This is because the binary representation of decimal numbers is not precise and you get rounding errors. This is the standard answer that most others will give you.

In this specific case, you don't actually need to compare floating-point numbers, because you're just testing whether some text represents a specific number. Since you're in shell, you can either use a regular string compare against "0.000" - assuming your data is rounded in that way - or using regular expressions with grep/egrep. Something like

egrep -q '0(|\.0+)'

Will match 0, 0.0, 0.00, etc, and will exit indicating success or failure, which you can use in the surrounding if statement:

if cut and pipe soup | egrep ... ; then
  ...
fi

Answer 2

Use a string comparison instead. Replace:

-eq 0

with:

= '0.000'

---

TZ:

Script section from comment:

for clus in $(ls *.cluster) ; do
    while read line ; do
        if [[ $(echo $line | cut -d" " -f11) -gt 0 ]] && [[ "$(echo $line | cut -d" " -f15 | bc -l)" = '0.000' ]] ; then
            cat $(echo $line | cut -d" " -f6).pdb >> test/$(echo $line | cut -d" " -f2)_pisa.pdb
        fi
    done < $clus
done

My pseudo-Python interpretation:

for clus in *.cluster:
    for line in clus:
        fields = line.split(' ')
        # field numbers are counting from 1 as in cut
        if int(field 11) > 0 and str(field 15) == '0.000':
            fin_name = (field 6) + '.pdb'
            fout_name = (field 2) + '_pisa.pdb'
            cat fin_name >> fout_name

Is that what you intended?