Remove elements from an array not containing a certain character

Question

I have an array c_arr containing table columns with table aliases. But there are some elements in the array which are actually not columns and so they don't have the format alias.column_name.

I need to remove those elements which do not contain a .. How can I do that?

The array is created using below statement:

c_arr=($echo $(grep -io "\b$alias.\w[a-zA-Z_0-9]*" $output_file))

There is another problem with the above line. Even though I am searching for $alias. (having a dot after the alias), the array c_arr is getting other values too which do not contain a dot. Sample values of the array are as follows:

cab.SYSTEM_NAME
cab.row_id
cab.name
cabxa
cabxa
cab.x_sys_name
cab.status_Cd
cab.LAST_UPD

UPDATE:

Now, the question at hand is how to remove the elements in the array c_arr which does not contain the character . if at all the array is having dot and non-dot elements. Contents of c_arr is as below:

cab.SYSTEM_NAME
cab.row_id
cab.name
cabxa
cabxa
cab.x_sys_name
cab.status_Cd
cab.LAST_UPD

The output desired is :

cab.SYSTEM_NAME
cab.row_id
cab.name
cab.x_sys_name
cab.status_Cd
cab.LAST_UPD

Answer 1

Then you would probably want to do something like this:

c_arr=$(echo $(grep -i "${alias}\." $output_file))

Because as Bill Jetzer mentioned you have to escape the dot in a regular expression.

If you are sure that everything with a dot is valid you could even do:

c_arr=$(echo $(grep -i \. $output_file))

Answer 2

Specifically answering the question "How to remove elements not containing a certain character from a Bash array". It's very similar in structure to Stéphane Chazelas' answer. If you have an array and you want to pare it down, you can do it as illustrated below:

$ list=(cab.row_id cab.name cabxa cabxa cab.x_sys_name)
$ printf "%s\n" ${list[@]}
cab.row_id
cab.name
cabxa
cabxa
cab.x_sys_name

$ # vvvv THIS ANSWERS THE QUESTION vvvv
$ IFS=$'\n' readarray -t list < <(printf "%s\n" ${list[@]} | grep '\.')
$ # ^^^^ THIS ANSWERS THE QUESTION ^^^^

$ printf "%s\n" ${list[@]}
cab.row_id
cab.name
cab.x_sys_name

This assumes that no elements contain a \n so it can be used as a delimiter.

Somewhat of a tangent: You can use unset to remove specific elements from an array, and that will cause Bash to skip over that element when iterating over the elements, but it will not cause all subsequent elements to shift their array index value:

$ list=(idx0 idx1 idx2 idx3 idx4)
$ printf "%s\n" ${list[@]}
idx0
idx1
idx2
idx3
idx4

$ unset list[2]
$ printf "%s\n" ${list[@]}
idx0
idx1
idx3
idx4

$ for x in ${list[@]}; do echo $x; done
idx0
idx1
idx3
idx4

$ for i in $(seq 0 4); do echo "$i: ${list[$i]}"; done
0: idx0
1: idx1
2: 
3: idx3
4: idx4

Answer 3

. is the regular expression operator that matches any single character.

Here, you'd want:

readarray -t c_arr < <(LC_ALL=C grep -iPo "\b\Q$alias\E\.\w+")

(assuming $alias doesn't contain \E and your grep supports -P (it already supports -o which is a GNU extension, so it's likely it will support -P as well)).