//! Search-path microbenchmarks.
//!
//! Every group pairs two ways of doing the same work in one run, so the
//! delta is a measurement rather than an estimate; a losing arm records
//! that an obvious-looking idea was measured and did not pay.
//!
//! Run with:
//!
//! ```text
//! cargo bench -p quicksearch-core --bench search
//! ```
//!
//! Sizes come from `corpus::SIZES`.

mod corpus;

use divan::Bencher;
use quicksearch_core::query::pattern::{TermPart, TermPattern};
use quicksearch_core::search::fuzzy::Bitap;
use quicksearch_core::snippet;

fn main() {
    divan::main();
}

fn literal(term: &str) -> TermPattern {
    TermPattern::build(&[TermPart {
        text: term.to_string(),
        glob: false,
    }])
    .expect("literal patterns always compile")
}

/// Decompressing the stored body, the first thing every full-text row does.
/// Reusing one context and pre-sizing the output is 5.1x faster at 1 KiB, the
/// size most documents are; `DocDecoder` exists for that.
mod zstd_decode {
    use super::*;

    #[divan::bench(args = corpus::SIZES)]
    fn decode_all(bencher: Bencher, size: usize) {
        let blob = corpus::blob(size, 4);
        bencher.bench(|| zstd::decode_all(divan::black_box(blob)).unwrap());
    }

    #[divan::bench(args = corpus::SIZES)]
    fn bulk_reused(bencher: Bencher, size: usize) {
        let blob = corpus::blob(size, 4);
        let capacity = corpus::text(size, 4).len();
        let mut dec = zstd::bulk::Decompressor::new().unwrap();
        bencher.bench_local(move || dec.decompress(divan::black_box(blob), capacity).unwrap());
    }
}

/// Turning decompressed bytes into a `&str`: `from_utf8_lossy` copies and
/// scan-validates; `String::from_utf8` vector-validates and *moves*. A
/// 14-16x gap; `DocDecoder` borrows instead.
mod utf8 {
    use super::*;

    #[divan::bench(args = corpus::SIZES)]
    fn lossy_into_owned(bencher: Bencher, size: usize) {
        let raw = corpus::text(size, 4).as_bytes();
        bencher.bench(|| String::from_utf8_lossy(divan::black_box(raw)).into_owned());
    }

    #[divan::bench(args = corpus::SIZES)]
    fn lossy_borrowed(bencher: Bencher, size: usize) {
        let raw = corpus::text(size, 4).as_bytes();
        bencher.bench(|| {
            let cow = String::from_utf8_lossy(divan::black_box(raw));
            cow.len()
        });
    }

    /// The decompressed `Vec` is unshared, so `from_utf8` can move it;
    /// the lossy fallback keeps corrupt-row behaviour unchanged.
    #[divan::bench(args = corpus::SIZES)]
    fn from_utf8_move(bencher: Bencher, size: usize) {
        let raw = corpus::text(size, 4).as_bytes();
        bencher
            .with_inputs(|| raw.to_vec())
            .bench_values(|owned| match String::from_utf8(owned) {
                Ok(s) => s,
                Err(e) => String::from_utf8_lossy(e.as_bytes()).into_owned(),
            });
    }
}

/// ASCII-folding the document. Folding into a reused buffer is not faster;
/// `fold_into` is chosen for what it does to the allocator, not the clock, so
/// do not "optimize" it back.
mod fold {
    use super::*;

    #[divan::bench(args = corpus::SIZES)]
    fn to_ascii_lowercase(bencher: Bencher, size: usize) {
        let text = corpus::text_mixed(size, 4);
        bencher.bench(|| divan::black_box(text).to_ascii_lowercase());
    }

    #[divan::bench(args = corpus::SIZES)]
    fn into_reused_buffer(bencher: Bencher, size: usize) {
        let text = corpus::text_mixed(size, 4);
        let mut buf = String::new();
        bencher.bench_local(move || {
            buf.clear();
            buf.push_str(divan::black_box(text));
            // `make_ascii_lowercase` preserves byte length, the invariant
            // the cascade's folded offsets rely on.
            buf.make_ascii_lowercase();
            buf.len()
        });
    }
}

/// Substring search: std's Two-Way against `memmem`'s SIMD prefilter. The
/// miss case matters most (a pass verifies far more rows than it accepts):
/// 92 µs against 4.7 µs at 256 KiB (20x). `match_indices` stays as the
/// guard: if the two converge, the SIMD path has stopped being selected.
mod substring {
    use super::*;

    #[divan::bench(args = corpus::SIZES)]
    fn match_indices_miss(bencher: Bencher, size: usize) {
        let text = corpus::text(size, 0);
        bencher.bench(|| divan::black_box(text).match_indices(corpus::NEEDLE).count());
    }

    #[divan::bench(args = corpus::SIZES)]
    fn memmem_miss(bencher: Bencher, size: usize) {
        let text = corpus::text(size, 0).as_bytes();
        let finder = memchr::memmem::Finder::new(corpus::NEEDLE);
        bencher.bench(|| finder.find_iter(divan::black_box(text)).count());
    }

    /// A `Finder` built per call: +14 ns against a hoisted one at 1 KiB and
    /// indistinguishable from 16 KiB up, since the precompute is O(needle)
    /// and a term is a few bytes.
    #[divan::bench(args = corpus::SIZES)]
    fn memmem_per_call_miss(bencher: Bencher, size: usize) {
        let text = corpus::text(size, 0).as_bytes();
        bencher.bench(|| {
            memchr::memmem::find_iter(divan::black_box(text), corpus::NEEDLE.as_bytes()).count()
        });
    }

    #[divan::bench(args = corpus::SIZES)]
    fn memmem_per_call_hits(bencher: Bencher, size: usize) {
        let text = corpus::text(size, 64).as_bytes();
        bencher.bench(|| {
            memchr::memmem::find_iter(divan::black_box(text), corpus::NEEDLE.as_bytes()).count()
        });
    }

    #[divan::bench(args = corpus::SIZES)]
    fn match_indices_hits(bencher: Bencher, size: usize) {
        let text = corpus::text(size, 64);
        bencher.bench(|| divan::black_box(text).match_indices(corpus::NEEDLE).count());
    }

    #[divan::bench(args = corpus::SIZES)]
    fn memmem_hits(bencher: Bencher, size: usize) {
        let text = corpus::text(size, 64).as_bytes();
        let finder = memchr::memmem::Finder::new(corpus::NEEDLE);
        bencher.bench(|| finder.find_iter(divan::black_box(text)).count());
    }

    /// What `pass_fulltext` runs per row on the literal path: one folded
    /// extraction yielding the window and both counts.
    #[divan::bench(args = corpus::SIZES)]
    fn cascade_row_sweeps(bencher: Bencher, size: usize) {
        let text = corpus::text_mixed(size, 4);
        let folded = corpus::text_folded(size, 4);
        let term = corpus::NEEDLE.to_ascii_lowercase();
        let budget = 600;
        bencher.bench(|| {
            snippet::extract_folded(
                divan::black_box(text),
                divan::black_box(folded),
                &term,
                corpus::NEEDLE,
                budget,
            )
            .map(|f| (f.count, f.exact, f.snippet.ranges.len()))
        });
    }

    /// The replaced shape: a separate case-sensitive count sweeps the
    /// document a second time for a number the extraction now has.
    #[divan::bench(args = corpus::SIZES)]
    fn cascade_row_sweeps_separate_count(bencher: Bencher, size: usize) {
        let text = corpus::text_mixed(size, 4);
        let folded = corpus::text_folded(size, 4);
        let term = corpus::NEEDLE.to_ascii_lowercase();
        let budget = 600;
        bencher.bench(|| {
            let exact =
                memchr::memmem::find_iter(divan::black_box(text).as_bytes(), corpus::NEEDLE)
                    .count();
            let f = snippet::extract_folded(text, folded, &term, corpus::NEEDLE, budget);
            (exact, f.map(|f| (f.count, f.snippet.ranges.len())))
        });
    }
}

/// Snippet extraction against a pre-folded haystack: the only folded sweep
/// of a row, and the source of its count.
mod snippet_extract {
    use super::*;

    #[divan::bench(args = corpus::SIZES)]
    fn extract_folded(bencher: Bencher, size: usize) {
        let text = corpus::text_mixed(size, 64);
        let folded = corpus::text_folded(size, 64);
        let term = corpus::NEEDLE.to_ascii_lowercase();
        let budget = 600;
        bencher.bench(|| {
            snippet::extract_folded(
                divan::black_box(text),
                divan::black_box(folded),
                &term,
                corpus::NEEDLE,
                budget,
            )
        });
    }
}

/// The filename pass's per-row ladder over 2000 realistic rows. `current` is
/// the shipped `find_first`, which jumps between spellings of the term's
/// first byte with `memchr2`. Adopting that took it from 53.4 µs to 45.4 µs,
/// 1.18x (2026-10-11, mimalloc, fastest sample, median of 3 runs), with the
/// other arms within 4%. Folding first does not catch up: the two obvious
/// fixes each measure *worse* alone (reused fold buffer, prebuilt `Finder`),
/// and together they only tie it. Short strings do not behave like document
/// bodies.
///
/// The other arms are local copies, which skip `find_first`'s per-call cost
/// (about 1.5 ns, 6 µs a run). The `memchr2` copy that prompted the change
/// measured 1.39x where the shipped call gained 1.18x, so gate a change on
/// `current` before and after it, not on a copy.
mod filename_ladder {
    use super::*;

    #[divan::bench]
    fn find_first_ci_current(bencher: Bencher) {
        let pattern = literal("quartzite");
        let rows = corpus::rows();
        bencher.bench(|| {
            let mut found = 0usize;
            for row in divan::black_box(rows) {
                if pattern.find_first(&row.name, true).is_some()
                    || pattern.find_first(&row.path, true).is_some()
                {
                    found += 1;
                }
            }
            found
        });
    }

    #[divan::bench]
    fn find_first_ci_scratch(bencher: Bencher) {
        let rows = corpus::rows();
        let mut scratch = String::new();
        bencher.bench_local(move || {
            let mut found = 0usize;
            for row in divan::black_box(rows) {
                scratch.clear();
                scratch.push_str(&row.name);
                scratch.make_ascii_lowercase();
                if memchr::memmem::find(scratch.as_bytes(), b"quartzite").is_some() {
                    found += 1;
                    continue;
                }
                scratch.clear();
                scratch.push_str(&row.path);
                scratch.make_ascii_lowercase();
                if memchr::memmem::find(scratch.as_bytes(), b"quartzite").is_some() {
                    found += 1;
                }
            }
            found
        });
    }

    /// Isolates the searcher from the allocation.
    #[divan::bench]
    fn find_first_ci_memmem(bencher: Bencher) {
        let finder = memchr::memmem::Finder::new("quartzite");
        let rows = corpus::rows();
        bencher.bench(|| {
            let mut found = 0usize;
            for row in divan::black_box(rows) {
                if finder
                    .find(row.name.to_ascii_lowercase().as_bytes())
                    .is_some()
                    || finder
                        .find(row.path.to_ascii_lowercase().as_bytes())
                        .is_some()
                {
                    found += 1;
                }
            }
            found
        });
    }

    /// Both at once: one reused fold buffer and a prebuilt `Finder`.
    #[divan::bench]
    fn find_first_ci_scratch_memmem(bencher: Bencher) {
        let finder = memchr::memmem::Finder::new("quartzite");
        let rows = corpus::rows();
        let mut scratch = String::new();
        bencher.bench_local(move || {
            let mut found = 0usize;
            for row in divan::black_box(rows) {
                scratch.clear();
                scratch.push_str(&row.name);
                scratch.make_ascii_lowercase();
                if finder.find(scratch.as_bytes()).is_some() {
                    found += 1;
                    continue;
                }
                scratch.clear();
                scratch.push_str(&row.path);
                scratch.make_ascii_lowercase();
                if finder.find(scratch.as_bytes()).is_some() {
                    found += 1;
                }
            }
            found
        });
    }

    /// The per-byte scan `find_first` ran until 2026-10-11: every start
    /// tested, `needle` already lowercased. The `memchr2` scan that replaced
    /// it was first rejected under glibc as not worth it on short haystacks;
    /// re-measured under mimalloc, its copy here ran 38.5 µs against
    /// `current`'s 53.4 µs. This copy runs 47.9 µs.
    #[divan::bench]
    fn find_first_ci_bytewise(bencher: Bencher) {
        let rows = corpus::rows();
        let needle = corpus::NEEDLE.as_bytes();
        fn find(hay: &[u8], needle: &[u8]) -> Option<usize> {
            let first = needle[0];
            let last = hay.len().checked_sub(needle.len())?;
            (0..=last).find(|&i| {
                hay[i].to_ascii_lowercase() == first
                    && hay[i..i + needle.len()].eq_ignore_ascii_case(needle)
            })
        }
        bencher.bench(|| {
            let mut found = 0usize;
            for row in divan::black_box(rows) {
                if find(row.name.as_bytes(), needle).is_some()
                    || find(row.path.as_bytes(), needle).is_some()
                {
                    found += 1;
                }
            }
            found
        });
    }
}

/// Bitap, the fuzzy passes' inner loop, run over every candidate row when
/// fuzzy is on. `step` takes `&mut [u64]`, so indices are bounds-checked and
/// the trip count is opaque to the optimizer.
mod bitap {
    use super::*;

    #[divan::bench(args = corpus::SIZES)]
    fn count_and_first_k2(bencher: Bencher, size: usize) {
        let bitap = Bitap::new(corpus::NEEDLE.as_bytes(), 2).unwrap();
        let hay = corpus::text(size, 4).as_bytes();
        bencher.bench(|| bitap.count_and_first(divan::black_box(hay)));
    }

    /// The filename pass's shape: many short haystacks, the register memset
    /// amortized over very little work.
    #[divan::bench]
    fn best_distance_over_names_k2(bencher: Bencher) {
        let bitap = Bitap::new(b"quartzite", 2).unwrap();
        let rows = corpus::rows();
        bencher.bench(|| {
            let mut hits = 0usize;
            for row in divan::black_box(rows) {
                if bitap.best_distance_and_first(row.name.as_bytes()).is_some() {
                    hits += 1;
                }
            }
            hits
        });
    }
}