Searching in bash shell - bash

Question

I have a text file.
Info in text file is
Book1:Author1:10.50:50:5
Book2:Author2:4.50:30:10
First one is book name, second is author name, third is the price, fourth is the quantity and fifth is the quantity sold.
Currently I have this set of codes
function search_book
{
read -p $'Title: ' Title
read -p $'Author: ' Author
if grep -Fq "${Title}:${Author}" BookDB.txt
then
record=grep -c "${Title}:${Author}" BookDB.txt
echo "Found '" $record "' record(s)"
else
echo "Book not found"
fi
}
for $record, I am trying the count the number of lines that is found. Did I do the right thing for it because when I run this code, it just shows error command -c.
When i did this
echo "Found"
grep -c "${Title}" BookDB.txt
echo "record(s)"
It worked, but the output is
Found
1
record(s)
I would like them to be together
Can I also add grep -i to grep -Fq in order to make all into small letters for better searching?
Lets say if I want to search Book1 and Author1, if I enter 'ok' for title and 'uth' for author, is there any % command to add to the title to search in the middle of the title and author?
The expected output is also expected to be..
Found 1 record(s)
Book1,Author1,$10.50,50,5.
Is there any where I can change the : delimiter to ,?
And also adding $ to the 3rd column which is the rice?
Please help..

Answer 1

Changing record=grep -c "${Title}:${Author}" BookDB.txt to record=$(grep -c "${Title}:${Author}" BookDB.txt) will fix the error. record=$(cmd) means assigning the output of command cmd to the variable record. Without that, shell will interpret record=grep -c ... as a command -c prepended by a environment variable setting(record=grep).
BTW, since your DB format is column-oriented text data, awk should be a better tool. Sample code:
function search_book
{
read -p $'Title: ' Title
read -p $'Author: ' Author
awk -F: '{if ($1 == "'"$Title"'" && $2 ~ "'"$Author"'") {count+=1; output=output "\n" $0} }
END {
if (count > 0) {print "found", count, "record(s)\n", output}
else {print "Book not found";}}' BookDB.txt
}
As you can see, using awk makes it easier to change delimiter(e.g. awk -F, for comma delimiter), and also makes the program more robust(e.g. it restricts the matching string to the first two fields). If you only need fuzzy match instead of exact match, you could change == to ~ in condition.

Answer 2

The "unnamed command -c" error can be avoided by enclosing the right part of the assignment in backticks or "$()", e.g.:
record=`grep -ic "${Title}:${Author}" BookDB.txt`
record=$(grep -ic "${Title}:${Author}" BookDB.txt)
Also, this snippet shows that -i is perfectly fine. However, please note that both grep commands should use the same list of flags (-F is missing in the 2nd one) - except for -q, of course.
Anyway, performing grep twice is probably not the best way to go. What about...
record=`grep -ic "${Title}:${Author}" BookDB.txt 2>/dev/null`
if [ ! -z "$record" ]; then ...
... or something like that?
By the way: If you omit -F you allow the user to operate with regular expressions. This would not only provide wildcards but also the possibility for more complex patterns. You could also apply an option to your script that decides whether to use -F or not..
Last but not least: To modify the lines, in order to change the delimiter or manipulate the columns at all, you could look into the manual pages or awk(1) or cut(1), at least. Although I believe that a more sophisticated language is more suitable here, e.g. perl(1) or python(1), especially when the script is to be extended with more features.

Answer 3

to add to the answer(s) above (this started as a comment, but it grew...) :
the $() form is preferred:
- it allows nesting,
- and it simplifies a lot the use of " and ' (each "level" of nesting see them at their level, so to speak). Tough to do with as using nested quotes and single-quotes becomes a nightmare of` and \\... depending on the "level of subshell" they are to be interpreted in...
ex: (trying to only grep once)
export results="$(grep -i "${Title}:${Author}" BookDB.txt)" ;
export nbresults=$(echo "${results}" | wc -l) ;
printf "Found %8s record(s)\n" "nbresults" ;
echo "$nbresults" ;
or, if too many results to fit in variable:
export tmpresults="/tmp/results.$$"
grep -i "${Title}:${Author}" BookDB.txt > "${tmpresults}"
export nbresults=$(wc -l "${tmpresults}") ;
printf "Found %8s record(s)\n" "nbresults" ;
cat "${tmpresults}" ;
rm -f "${tmpresults}" ;
Note: I use " a lot (except on the wc -l line) to illustrate it could be needed sometimes (not in all the cases above!) to keep spaces, newlines, etc. (And I purposely drop it for nbresults so that it only contain the number of lines, not the preceding spaces).