Stream: beginners

Topic: ✔ Is pattern matching equivalent to `if x == y`?

Felipe Vogel (Sep 11 2024 at 19:35):

In the tutorial section on pattern matching, this snippet:

if stoplightColor == Red then
    "red"
# ...

is converted to:

when stoplightColor is
    Red -> "red"
# ...

Does that mean pattern matching is equivalent to if x == y?

That seems to be the case as long as the matches are a single value. This works:

when 1 + 1 is
    2 -> Stdout.line! "math is cool"
    _ -> crash "everything you believe in is a lie"

But this doesn't work, with the match as an expression:

when 1 + 1 is
    3 - 1 -> Stdout.line! "math is cool"
    _ -> crash "everything you believe in is a lie"

I get this error:

MISSING ARROW

I am partway through parsing a `when` expression, but got
stuck here:

6│  when 1 + 1 is
7│      3 - 1 ->
          ^

I tried adding parentheses around the expression, and putting the branches on a new indented line, but the error is still there. It seems to be related to the way pattern matching works. Which makes me wonder: is pattern matching really equivalent to if x == y after all?

Sam Mohr (Sep 11 2024 at 19:38):

It's similar, but not equivalent. We call it matching instead of equivalence checking because we're checking if a value with different possible states matches a pattern that represents a set of those states.

For example,

when user is
    { email, status: Inactive } ->
        Stdout.line! "Inactive user found with email $(email)"

Sam Mohr (Sep 11 2024 at 19:39):

This matches on any user that has status == Inactive, but will allow any user with any email.

Sam Mohr (Sep 11 2024 at 19:41):

The examples you gave at the start are these special cases where instead of matching on structurally equivalent values like { email, status: ... }, we're instead matching on something that has the same whole value, like 2.

Sam Mohr (Sep 11 2024 at 19:41):

The 3 - 1 thing doesn't work because we don't currently (nor do most languages) have a compiler that's smart enough to know that 3 - 1 is equivalent to 2.

Sam Mohr (Sep 11 2024 at 19:43):

So in short, when is close to ==, but it's smarter in the way that it can capture and use structural content of the matched data, but dumber in that it needs to be given static values, not expressions that evaluate to static values

Felipe Vogel (Sep 11 2024 at 19:54):

Gotcha, that makes sense, thanks.

The reason I went down this rabbit hole is that _ if someExpression -> match syntax makes it easier to fit logic into pattern matching that otherwise wouldn't fit and would have to be changed to a full if-then expression instead. So that made me wonder about the boundaries between the two syntaxes.

Sam Mohr (Sep 11 2024 at 19:59):

Well, I disagree that a single if expression linguistically implies you're checking multiple conditions. That's why we have these if-else chains:

"If X then A else if Y then B else if ..."

when is meant to imply you're checking multiple conditions. If we found a word that implied both, I'd be more okay with them having the same keyword, but if is so universal, it'd be pretty hard to argue for getting rid of it

Felipe Vogel (Sep 11 2024 at 20:13):

Sorry, I didn't mean to suggest changing when to if. I meant that I was thinking about the existing way that if can be included in a when match:

when n is
    0 -> # ...
    1 -> # ...
    _ if Num.isEven n -> # ...
    _ -> # ...

(And the reason I was thinking about it is that this structure appears in some solutions to the Roc Collatz Conjecture exercise on Exercism.)

Sam Mohr (Sep 11 2024 at 20:14):

Oh, I see. Those are guard statements

Sam Mohr (Sep 11 2024 at 20:14):

Yeah, let me re-read your messages, I may have read too quickly

Sam Mohr (Sep 11 2024 at 20:17):

Okay, having re-read, I think the reason I answered the way I did was because you ended on "that made me wonder about the boundaries between the two syntaxes", and I didn't know how to provide a useful answer to that, so I assumed you were implying the barrier should be dropped.

Sam Mohr (Sep 11 2024 at 20:18):

But I'm not actually sure what the question was, unless you really were just offering food for thought to the group.

Sam Mohr (Sep 11 2024 at 20:18):

Which is totally fine! I'm just tutor-brained right now

Felipe Vogel (Sep 11 2024 at 20:22):

No worries, you answered my question when you explained:

The 3 - 1 thing doesn't work because we don't currently (nor do most languages) have a compiler that's smart enough to know that 3 - 1 is equivalent to 2.

When I said "that made me wonder…" I was just saying why I asked the question in the first place, because I didn't want to seem like I was arbitrarily complaining about an edge case that doesn't matter in any real scenarios. My mistake for not being clearer that you'd already answered my question! Thanks again.

Sam Mohr (Sep 11 2024 at 20:22):

Great!

Notification Bot (Sep 17 2024 at 18:38):

Felipe Vogel has marked this topic as resolved.

Last updated: Jul 06 2025 at 12:14 UTC