Generate ID number from a name in bash - bash

Question

Currently I have a bunch of names that are tied to numbers, for example:
Joe Bloggs - 17
John Smith - 23
Paul Smith - 24
Joe Bloggs - 32
Using the name and the number I'd like to generate a random/unique ID made of 4 numbers that also ends with the initial number.
So for example, Joe Bloggs and 17 would make something random/unique like: xxxx17.
Is this possible in bash? Would it be better in some other language?
This would be used on debian and darwin based systems.

Answer 1

It is impossible to ensure than 4-digit hash (checksum) would be unique for a set of 10 character long names.
As an alternative, you can try
file="./somefile"
paste -d"\0\n" <(seq -f "%04g" 9999 | sort -R | head -$(grep -c '' "$file")) <(grep -oP '\d+' "$file")
for better readability
paste -d"\0\n" <(
seq -f "%04g" 9999 | gsort -R | head -$(grep -c '' "$file")
) <(
grep -oP '\d+' "$file"
)
for your input produces something like:
010817
161523
748024
269032
All lines are in the form RRRRXX, where:
the RRRR is an guaranteed unique and random number (from the range 0001 up to 9999)
the XX is the number from your input
decomposition:
seq produces 9999 4-digit numbers (ofc, each number is unique)
sort -R sorts the lines in random order (based on their hash, so get unique random numbers)
head - from the random list show only first N lines, where the N is the number of lines in your file,
the number of lines is counted by grep -c '' (better than wc -l)
the grep -oP filters the numbers from your file
finally the paste combines the two inputs to the final output
the <(..) <(..) is process substitution

Answer 2

Each name, after you add their number, becomes unique already unless there are two Joe Bloggs 17. In your case, there are two Joe Bloggs, one with 17 and 32. Put those together, you have uniqueness "Joe Bloggs 17" and "Joe Bloggs 32" are not the same. Using this, you can simply assign a number to each name + number pair and remember that number in an associative array (dictionary). No need to be random. When you find a name that isn't already in the dictionary, just keep incrementing the number and, then, associate the new number with the name. If uniqueness is the only goal, then you are in good shape for 10,000 people.
Python is a great language for this, but you can make associative arrays in BASH too.

Answer 3

You can get very close to doing exactly what you want using the random string generated by $(date +%N) and then selecting 4 digits to use as the first for characters in the new ID. You can choose from the beginning if you want IDs that are closer together, or from the mid part of the string for more randomness. After selecting your random 4, then just keep track of the ones used in an array and check against the array as each new ID is assigned. This overhead is negligible for 10,000 or so IDs:
#!/bin/bash
declare -a used4=0 # array to hold IDs you have assigned
declare -i dupid=0 # a flag to prompt regeneration in case of a dup
while read -r line || [ -n "$line" ]; do
name=${line% -*}
id2=${line##* }
while [ $dupid -eq 0 ]; do
ns=$(date +%N) # fill variable with nanoseconds
fouri=${ns:4:4} # take 4 integers (mid 4 for better randomness)
# test for duplicate (this is BASH only test - use loop if portability needed)
[[ "$fouri" =~ "${used4[#]}" ]] && continue
newid="${fouri}${id2}" # contatinate 4ints + orig 2 digit id
used4+=( "$fouri" ) # add 4ints to used4 array
dupid=1
done
dupid=0 # reset flag
printf "%s => %s\n" "$line" "$newid"
done<"$1"
output:
$ bash fourid.sh dat/nameid.dat
Joe Bloggs - 17 => 762117
John Smith - 23 => 603623
Paul Smith - 24 => 210424
Joe Bloggs - 32 => 504732