Regex using Bash metacharacters in double bracket `[[…]]` command using `=~` operator

Question

The Bash reference manual, in sec. 3.2.5.2 Conditional Constructs, says

It is sometimes difficult to specify a regular expression properly without using quotes, or to keep track of the quoting used by regular expressions while paying attention to shell quoting and the shell’s quote removal.

So I started playing with it.

[[ ' 2' =~ 1 | 2 ]] && echo YES || echo NO

gives the error

bash: syntax error in conditional expression: unexpected token `|'
bash: syntax error near `|'

Also

[[ ' 2' =~ 1 \| 2 ]] && echo YES || echo NO

gives the error

bash: syntax error in conditional expression
bash: syntax error near `|'

My question is: why the above do not work, considering the following?

| is a bash metacharacter and also a regular expression special character.

"The words between the [[ and ]] do not undergo word splitting and filename expansion. The shell performs tilde expansion, parameter and variable expansion, arithmetic expansion, command substitution, process substitution, and quote removal on those words (the expansions that would occur if the words were enclosed in double quotes)". Ref. Bash reference manual sec. 3.2.5.2 Conditional Constructs).

So I expect Bash consider ' 2' =~ 1 | 2 as a single word, in particular not splitting at bash metacharacter |.

When you use ‘=~’, the string to the right of the operator is considered a POSIX extended regular expression pattern and matched accordingly. Ref. Bash reference manual sec. 3.2.5.2 Conditional Constructs).

So I expect Bash to interpret 1 | 2 as a regex.

Even if | is parsed as a metacharacter by bash, I expect that quoting it (like in \|) would be good as Bash would pass 1 | 2 to the regex engine.

I know how to make it work, for example

[[ ' 2' =~ '1 '|' 2' ]] && echo YES || echo NO

prints 'YES'.

Answer 1

If you really want to go to escaping route then you will need to escape all the whitespaces in your regex to tell shell that it is a single word like this:

[[ ' 2' =~ 1\ |\ 2 ]] && echo YES || echo NO
YES

However cleaner approach is to store regex in a variable and then use it like this:

rx='1 | 2'
[[ ' 2' =~ $rx ]] && echo YES || echo NO
YES

Answer 2

As a variation to the answer given by @anubhava :

I find escaping less pleasant than quoting, and I would write something like

[[ ' 2' =~ "1 "|" 2" ]]

Note that you can not quote the whole expression, because everything between quotes is taken literally, not as regex. Therefore the | must be outside the quotes.