lexer returns SyntaxError for i64::MIN literal

[hnakamur]

Currently the lexer returns SyntaxError for i64::MIN literal {{ -9223372036854775808 }}

I added it to tests/lexer-inputs/literals.txt
hnakamur@4cbcac8
and I got the following error.

thread 'test_lexer' panicked at 'called `Result::unwrap()` on an `Err` value: Error { kind: SyntaxError, detail: "invalid integer" }', minijinja/tests/test_lexer.rs:17:41

The lexer parses an integer literal to i64 at

minijinja/minijinja/src/compiler/lexer.rs

Line 255 in 27413fb

.map(Token::Int)

However the negative sign and 9223372036854775808 are tokenized separately, so 9223372036854775808 overflows since 9223372036854775808 > i64::MAX (= 9223372036854775807).

I think it would be nice for the lexer to handle a negative integer literal and handle i64::MIN literal, or document the current behavior that the available integer literal is between i64::MIN + 1 and i64::MAX.

[mitsuhiko]

I wanted to fix up some of those edge cases for a while but I was too lazy. The correct thing to do would be for the lexer to have to tokens: a u64 for integers and a u128 for large integers. Then neg (which is currently invoked for all negative integers) gets a special case for 170141183460469231731687303715884105728 which is the largest possible positive integer that can still be negated. Independently of that, it's a bit ridiculous that the compiler emits opcodes for negations of numbers to begin with. It should really compile -42 into just a constant rather than Neg + LoadConst.

I will fix this.

[mitsuhiko]

#353

[hnakamur]

Thanks!