String concatenation doesn't work in bash - bash

Question

I am coding a shell script that reads text files and creates JSON key-value pairs based on them. The key is the filename and the value is a random line of the file.
The trouble is when I concatenate the key with the value in the global variable data.
When I run the code bellow:
data='{'
for file in $(ls _params)
do
key=${file%.txt}
f_line=$(($$%100))
value=$(sed "${f_line}q;d" "./_params/$file")
# assembles key-value pairs
data=$data\"$key\":\""value"\",
done
data=${data%?} # removes last comma
data="$data}"
echo $data
My output is: {"firstName":"value","lastName":"value"}
But changing the string "value" to the variable $value, as follows:
data='{'
for file in $(ls _params)
do
key=${file%.txt}
f_line=$(($$%100))
value=$(sed "${f_line}q;d" "./_params/$file")
# assembles key-value pairs
data=$data\"$key\":\"$value\",
done
data=${data%?} # removes last comma
data="$data}"
echo $data
The output gets confused: "}"lastName":"Adailee.
I wish to store in the $data variable something like: {"firstName":"Bettye","lastName":"Allison"}
Note: My bash version is 4.3.48.
Note: Inside my archive _params I have two files firstName.txt and lastName.txt both with random names each line.

Answer 1

As #ruakh suggests, the specific issue is your input files. Here are steps to repro your issue and verify this:
I created two a firstNames.txt file with A B C D repeated 100 times:
$ cat ABCD
A
B
C
D
$ for _ in $(seq 1 100); do cat ABCD >> _params/firstName.txt
And then similar with W X Y Z for lastNames.txt. Then I ran your script:
$ bash q.sh
{"firstName":"A","lastName":"W"
However, if I use unix2dos (from the dos2unix package) to convert this to \r\n line endings.
$ unix2dos _params/firstName.txt
unix2dos: converting file _params/firstName.txt to DOS format...
$ unix2dos _params/lastName.txt
unix2dos: converting file _params/lastName.txt to DOS format...
$ bash q.sh
"}"lastName":"W
So you could probably use dos2unix to fix your input files (or open vim and do :set ft=unix and then :x).
But I wanted to let you know about three other things.
$$ is not a random number, it's the PID of your current process.
best practice is not to parse ls, but to use globbing instead1
you can solve the fencepost problem without removing the comma you just placed by starting with the empty separator and setting it to comma after the first iteration of the loop.
Here is my suggestion for improving your script (once you fix the newlines in the input):
#!/bin/bash
data='{'
sep=""
for file in _params/*
do
key=${file%.txt}
file_length=$(wc -l < ${file})
f_line=$(( (RANDOM % file_length) + 1 ))
value=$(sed "${f_line}q;d" "${file}")
# assembles key-value pairs
data="${data}${sep} \"$key\":\"$value\""
sep=","
done
data="${data} }"
echo $data

Answer 2

$value apparently ends with a carriage return character (\r, U+000D). As a result, when you print it, the cursor moves back to the beginning of the line, and subsequent characters are printed starting at the first column, overwriting what was there before. (This doesn't affect the actual order of characters, of course; it's just displayed confusingly when you print it.)
To fix this, you can write
value="${value%$'\r'}"
to remove the trailing carriage return.