From: Junio C Hamano <gitster@pobox.com>
Date: Fri, 28 Jun 2019 18:05:03 GMT
Subject: Re: [PATCH 1/6] test-lib: introduce test_commit_bulk
Message-ID: <xmqqh8898u28.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <2d4410a9-fd3e-8b9f-00b5-f8eba4d51b42@gmail.com>

Derrick Stolee <stolee@gmail.com> writes:

> way to do this loop. The top thing on my mind are the 'eval "echo X"'
> lines. If they start processes, then we can improve the performance.
> If not, then it may not be worth it.

Sigh.  

Do you mean 'echo' run inside 'eval' is one extra process?  In most
modern shells, it is a built-in and you need another process.

Do you mean 'eval' running anything is one extra process?  Because
anything done inside eval must be visible to the shell running it,
e.g.

	var=myvar; eval "$var=val"

would evaluate string 'myvar=val' inside that shell itself and it
must be able to update the value of $myvar, whatever it does must
not add any extra process.

The primary reason why the loop in question uses eval is to allow
the callers to pass $n in single-quote to have it interpolated
lazily.

	message='message $n'
	for n in 1 2 3
	do
		echo "$message"
		eval "echo \"$message\""
	done

Each iteration, the first line gives

	message $n

which is the thing that gets passed to 'echo' in the second line, so
you'll see

	message $n
	message 1
	message $n
	message 2
	message $n
	message 3

as the result.