if-let behavior appears different in #1188

[sogaiu]

The following code used to not error for me:

(if-let [[l r] (peg/match "a" "b")]
  :a
  :b)

With 2c3ca29 I get:

$ janet
Janet 1.28.0-dev-2c3ca298 linux/x64/gcc - '(doc)' for help
repl:1:> (if-let [[l r] (peg/match "a" "b")]
repl:2:(>   :a
repl:3:(>   :b)
error: expected string, symbol, keyword, array, tuple, table, struct or buffer, got nil
  in _thunk [repl] (tailcall) on line 1, column 1

With 94722e5 I get:

$ ./build/janet
Janet 1.28.0-dev-94722e56 linux/x64/gcc - '(doc)' for help
repl:1:> (if-let [[l r] (peg/match "a" "b")]
repl:2:(>   :a
repl:3:(>   :b)
:b

[sogaiu]

Note that when-let (as it's implemented in terms of if-let) is also affected as well:

$ janet
Janet 1.28.0-dev-2c3ca298 linux/x64/gcc - '(doc)' for help
repl:1:> (when-let [[a b] nil] :a)
error: expected string, symbol, keyword, array, tuple, table, struct or buffer, got nil
  in _thunk [repl] (tailcall) on line 1, column 1

For comparison its older behavior was:

$ ./build/janet
Janet 1.28.0-dev-94722e56 linux/x64/gcc - '(doc)' for help
repl:1:> (when-let [[a b] nil] :a)
nil

[sogaiu]

I'm pretty sure I came to rely on the not-erroring behavior because it seemed more convenient. I didn't look into why it was working and I didn't check that this wasn't intended (at least I didn't look closely at various docstrings nor did I ask around).

I suppose if I'm alone in having gone this path then the good news is that there's only one person who needs to rewrite code :)

[primo-ppcg]

I suppose if I'm alone in having gone this path then the good news is that there's only one person who needs to rewrite code :)

(if-let [[a b] (possibly-nil-or-tuple)] ...)

Seems like a perfectly valid use case, at least to me. I'll make an amendment.

[zevv]

But this will mean the rules for not being to able to deconstruct a nil will be different from not being able to deconstruct something else?

(if-let [a b] (something-that-can-not-be-deconstructed-into-a-b-tuple) ...)

[primo-ppcg]

But this will mean the rules for not being to able to deconstruct a nil will be different from not being able to deconstruct something else?

Yes, by definition really.

if-let
If any are false or nil, evaluate the fal form.