Question

As bash doesn’t have first class functions, I’m simulating anonymous functions by passing a string to a function, which then gets evaluated by eval.

it() {
    echo "$1" # prints the description of the spec
    beforeEach # a setup function
    eval "$2"

    if (($? == 0)); then
        # do something
    fi

    afterEach  # a cleanup function
}

it "should echo something" '{
  echo "something"
}'

This allows to write very concise tests (it defines a specification). Now I wonder if this is a valid use of eval.

EDIT

I am aware that the opening { and closing } are not need in the anonymous function string, it’s just that that way it resembles for something like Jasmine.

EDIT

The pseudo anonymous function is actually a test, which means at the end of the test, there is something like

[[ var == 'foo' ]]

or

((i > 10))

i.e. some sort of test (or assert, in XUnit terms). It never needs to return anything than the return code which then gets evaluated, and if the return code is 0 (success), the description is printed in green, otherwise red (the test failed).

Answer 1

If it gets the job done and you (and ideally others) understand how it works, absolutely.

I’ve been writing similar code for test case automation more than once or twice, but I admit it can get pretty hairy, especially if you start simple and then allow it to grow organically.

For what it’s worth, I would like to suggest a stylistic improvement;

if eval "$2"; then
  # something

or if the “something” is a simple command, plain old

eval "$2" && # something

Having said that, it might be better if you can avoid the eval altogether, but this depends on your test cases.

$2 && # something