Re: i vote for $( .. <file ) be available also as non first command in subshells ( and maybe process subst ) ?

[Greg Wooledge]

On Fri, Mar 12, 2021 at 10:17:58AM +0100, Alex fxmbsw7 Ratchev wrote:
> .. :)

Please don't put the content of your message in the Subject: header.
The Subject: header may be lost, changed, truncated, ignored, etc.

In the interest of readability, let's copy it down here into the body:

> i vote for $( .. <file ) be available also as non first command
> in subshells ( and maybe process subst ) ?

OK, first things first: $(<file) is special syntax.  It is defined to
be equivalent to $(cat file), except that it doesn't need to exec cat(1),
so it is marginally more efficient.

As you observed in IRC, however, it is not the same:

x=$(cat nosuchfile 2>/dev/null)
x=$(<nosuchfile 2>/dev/null)

These two commands are not equivalent.  In the first one, the error message
is suppressed.  In the second, it is printed.  For some reason, you
thought this was important enough to complain about.

If you don't like the behavior of $(<nosuchfile) then simply don't use
it.  Use cat instead.  Or wrap the assignment in a curly-brace block
and move the 2>/dev/null.

{ x=$(<nosuchfile); } 2>/dev/null

All of them fork a new process, so the $(<) form is only a *tiny* bit
more efficient, not a lot more efficient.  You can simply stop using it.

Second thing second: outside of the special $(< syntax, <file is a
redirection.  If you write a redirection without a command, the behavior
is well known.  A standalone output redirection is used to truncate a
file -- the shell opens the file for writing (which truncates it), does
nothing, and then closes the file.  People use this all the time.

A standalone input redirection doesn't really serve any purpose.  The
shell will open the file for reading, do nothing, and then close the
file.  I suppose you could use this to verify that a file is truly
readable, but most people would simply use test -r.

Third thing third: $( ) is a command substitution.  Not a subshell, and
not a process substitution.  A subshell is ( ), and a process
substitution is <( ) or >( ).

(Yes, technically, all three of them create subshells, but come on, you
KNOW that an explicit subshell is ( ) not $( ).  You know better.  Stop
using the wrong names for things.)

> the story is for me i think, $( < file ) is useful, more useful advanced
> core even more useful, but then i cant do <file anymore to print its output
> which would also be handy in <( >( ..

I have no idea what "more useful advanced core even more useful" means.

Apart from  x=$(<nosuchfile 2>/dev/null)  which you didn't even mention
in the email -- only in IRC -- what exactly are you trying to do?  Show
us an example or something.

One of the other things you showed in IRC (why oh why can't you show them
in your email???) was this:

x=$(foo; bar; <file)

Apparently you thought this would be useful for some reason.  However,
this is never going to be equivalent to

x=$(foo; bar; cat file)

As mentioned previously, a standalone redirection already has a meaning.
You can't just "vote" to redefine it to mean "pretend I wrote cat".  The
special form $(<file) is special.  It's a one-time-only thing.  It's
a new piece of syntax that consists of three characters, but you're allowed
to put whitespace between the second and third character.

You can't just use a PIECE of this syntax in a different context.

If you don't have the three characters $(< in a row (with optonal whitespace
allowed in ONE position only), then you are not using this feature.  You
are doing something else.