Re: Parsing, metacharacters, tokens.

[Greg Wooledge]

On Thu, May 19, 2022 at 12:32:59PM +0200, Cristian Zoicas wrote:
> There are situations when it is not clear from the manual how bash
> performs expansions/splitting/substitutions.

Yeah.  Bash syntax is a nightmare.

> According to what is written in the manual a metacharacter is:
> 
>     A character that, when unquoted, separates words. A metacharacter
>     is one of the following characters:
>         <space> <tab> <newline> | & ; ( ) < >
> 
> Now let's assume the following statement:
> 
> A=START$(echo _)END
> 
> According to the definition of metacharacters the first token would be
> "A=START$" (because '(' separates tokens) and then will follow the rest of
> the tokens .

You're reading the quoted sentences far too literally, or else you're
adding some of your own assumptions to it.  I'm not sure which.

( is a metacharacter, but not when it's preceded by $.  There are various
combinations of characters that act like a single character, as far as
parsing goes.  $( is the introduction of a command substition, and it
doesn't cause a word separation the way that ( does.

Similarly, <& and >& act like a single metacharacter, not two different
metacharacters back to back.

In your example, START$(echo _)END is all one word.  The $(...) bit
will undergo a command substitution.  Then, because the word is on the
right hand side of a simple assignment, there WON'T be an additional
word-splitting round, despite the lack of explicit double quotes.