Bash: T9 simulator (intuitive text phone)

What bothers me is the structure of the bash... I’ll give you my algorithm in natural language so you can understand my reasoning.



Start algo

We ask the user to choose between converting number to letter or letter to number

case 1: number to letter


We ask the user to enter their sequence of numbers
then for each number we associate the group of letters (2 = (a/b/c); 3 = (d/e/f)...) which we display on the screen, then we look for an existing word in the file dico.txt with that combination

case 2: letter to number

We ask the user to enter their string of characters
then for each letter we associate the corresponding number (a = 2; b = 2, c = 2, d = 3...) which we display on the screen


end algo

So, for case 1, I stick to the table, which, however, will contain as many expressions in brackets (cf. man 7 regex) as there are digits available on the keyboard:

exprCroc=( [2]='[abc]' [3]='[def]' ...)

the search in dico.txt will simply be done with a

grep

using the regex pattern concatenated from the data provided by the user and retrieved from the table.
if the user provides:

23

, then the constructed regex will be

"^[abc][def]$"

.

in both cases, the provided data must be read one character at a time.

while read -n1 l; do test -n "$l" && regex+="$l"; done <<<"mon_mot"

for case 1, we will look for the corresponding value in the table.
for case 2, we will use an indirection (cf. man bash).