Some performance issues

[dalance]

I found some small performance issues.

disable log

if parol_runtime provide log/release_max_level_off feature like below, Rust compiler can remove trace! macro statically.

[features]
max_level_off   = ["log/max_level_off"]
max_level_error = ["log/max_level_error"]
max_level_warn  = ["log/max_level_warn"]
max_level_info  = ["log/max_level_info"]
max_level_debug = ["log/max_level_debug"]
max_level_trace = ["log/max_level_trace"]
release_max_level_off   = ["log/release_max_level_off"]
release_max_level_error = ["log/release_max_level_error"]
release_max_level_warn  = ["log/release_max_level_warn"]
release_max_level_info  = ["log/release_max_level_info"]
release_max_level_debug = ["log/release_max_level_debug"]
release_max_level_trace = ["log/release_max_level_trace"]

I got 6% performance gain.

throughput/parse        time:   [11.043 ms 11.056 ms 11.073 ms]
                        thrpt:  [1.3537 MiB/s 1.3558 MiB/s 1.3574 MiB/s]
                 change:
                        time:   [-7.8789% -5.9537% -4.3499%] (p = 0.00 < 0.05)
                        thrpt:  [+4.5478% +6.3307% +8.5528%]
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  2 (2.00%) high mild
  4 (4.00%) high severe

remove is_empty before pop

VerylGrammarAuto::pop which is generated by parol has is_empty check.
It can be removed because pop will return None when the vec is empty.

fn pop(&mut self, context: &str) -> Option<ASTType> {
    //if !self.item_stack.is_empty() {
    let item = self.item_stack.pop();
    if let Some(ref item) = item {
        trace!("pop     {}: {:?}", context, item);
    }
    item
    //} else {
    //    None
    //}
}

I got 2% performance gain.

throughput/parse        time:   [10.867 ms 10.931 ms 11.023 ms]
                        thrpt:  [1.3599 MiB/s 1.3713 MiB/s 1.3793 MiB/s]
                 change:
                        time:   [-4.6804% -2.4878% -0.5493%] (p = 0.01 < 0.05)
                        thrpt:  [+0.5523% +2.5513% +4.9102%]
                        Change within noise threshold.
Found 14 outliers among 100 measurements (14.00%)
  5 (5.00%) high mild
  9 (9.00%) high severe

[jsinger67]

Thanks for your analysis.
Both are good points 👍
I will incorporate your suggestions in the next version.

[jsinger67]

Fixed with commit f1661db.
New versions have not yet been released on crates.io.