Files
five/hbrtl/sqlscan.go
CharlesKWON 5c067f35a4 perf(hbrt): ExecPcodeFast — pcode variant without defer/recover
Pcode expressions compiled from SQL WHERE clauses (via genpc.CompileExpr)
never contain BEGIN SEQUENCE and can't raise BreakValue, so the defer +
recover dance in ExecPcode's EndProc is pure overhead. For FiveSql2's
per-row WHERE evaluation on a 50k-row scan, that's 50k × ~15ns = ~750µs
of pointless recover bookkeeping.

Split ExecPcode into two variants sharing execPcodeBody:

  ExecPcode     — full: Frame + defer EndProc. General-purpose,
                  handles panics. Behavior unchanged.

  ExecPcodeFast — hot: Frame + execPcodeBody + EndProcFast. No defer,
                  no recover. Caller guarantees the pcode body can't
                  panic with HbError / BreakValue.

SqlScan now uses ExecPcodeFast for per-row WHERE evaluation. Measured
impact on 50k-row no-WHERE benchmark: 10.6ms → 9.2ms steady state
(~13% faster). Effect is smaller on numeric-WHERE because per-row
cost there is dominated by the opcode dispatch itself, not the frame
exit.

Validation:
  - FiveSql2 43/43
  - go test ./hbrt/... PASS (pcode tests)
  - go test ./hbrtl/... PASS

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-14 12:07:54 +09:00

156 lines
4.7 KiB
Go

// Copyright (c) 2026 Charles KWON OhJun (charleskwonohjun@gmail.com)
// All rights reserved.
// Go-native SQL scan loop for FiveSql2 hot path.
//
// Motivation: FiveSql2 is a PRG-based SQL interpreter. For simple
// "SELECT cols FROM table WHERE cond" queries, the per-row cost is
// dominated by PRG interpreter overhead (AST tree walk, field name
// lookup, workarea switching). Moving just the inner scan loop to Go
// bypasses all that overhead and gets us ~15x speedup for the common
// case while keeping the rest of FiveSql2 untouched.
//
// The SQL engine remains responsible for:
// - Parsing SQL and building AST
// - Resolving field names to positions (column binding)
// - Compiling WHERE expression to pcode (via PcCompile)
// - GROUP BY, ORDER BY, aggregates (not per-row)
//
// This helper only handles the hot loop:
// - Full table scan (workarea already positioned)
// - Per-row WHERE evaluation via ExecPcode
// - Column extraction via cached field positions
// - Result array construction
package hbrtl
import (
"five/hbrdd"
"five/hbrt"
)
// SqlScan(aFieldPositions, pcWhere) → aRows
//
// Scans the current workarea top-to-bottom, evaluates pcWhere per row
// (nil = no filter), collects selected column values into rows.
//
// aFieldPositions: array of 1-based field positions to extract per row.
// Resolve once before calling (FieldPos cache is O(1)
// but still has PRG → Go call overhead).
// pcWhere: pcode function pointer from PcCompile, or NIL.
//
// Returns:
// Array of rows, each row = Array of field values.
//
// Notes on CHAR trimming: DBF character fields are space-padded. The
// caller decides whether to trim (via a SELECT-list AllTrim wrapper).
// We don't trim here — that's a semantic choice, and callers who need
// raw bytes shouldn't pay for a strings.TrimSpace().
func SqlScan(t *hbrt.Thread) {
t.Frame(2, 0)
defer t.EndProc()
// Parse arguments
fieldsVal := t.Local(1)
if !fieldsVal.IsArray() {
t.PushValue(hbrt.MakeArray(0))
t.RetValue()
return
}
fieldsArr := fieldsVal.AsArray().Items
nFields := len(fieldsArr)
whereVal := t.Local(2)
var whereFn *hbrt.PcodeFunc
if !whereVal.IsNil() {
if p := whereVal.AsPointer(); p != nil {
whereFn, _ = p.(*hbrt.PcodeFunc)
}
}
// Pre-convert field positions to []int (avoid Value->int per row)
fieldPos := make([]int, nFields)
for i := 0; i < nFields; i++ {
fieldPos[i] = int(fieldsArr[i].AsNumInt())
if fieldPos[i] < 1 {
fieldPos[i] = 1
}
}
wam, ok := t.WA.(*hbrdd.WorkAreaManager)
if !ok {
t.PushValue(hbrt.MakeArray(0))
t.RetValue()
return
}
area := wam.Current()
if area == nil {
t.PushValue(hbrt.MakeArray(0))
t.RetValue()
return
}
// SQLite-inspired: instead of one slice allocation per row, maintain
// a single flat backing buffer and hand each row a sub-slice into it.
// This halves allocations (row header + backing → just row header)
// and keeps row data contiguous in memory for better cache locality.
//
// Safety: we cap each sub-slice to exactly nFields via the 3-index
// slice form (flat[off:end:end]). Any later `append` on an individual
// row will then trigger a reallocation of that row's backing, so we
// don't clobber neighboring rows if PRG code mutates via AAdd.
// Size the initial backing based on the workarea's record count —
// even if WHERE filters most rows out, over-allocating beats five
// regrowths of a 200 KB buffer mid-scan.
estRows := 1024
if rc, err := area.RecCount(); err == nil && rc > 0 {
estRows = int(rc)
if estRows > 1 << 20 {
estRows = 1 << 20
}
}
rows := make([]hbrt.Value, 0, estRows)
flat := make([]hbrt.Value, 0, estRows*nFields)
slab := hbrt.NewArraySlab(estRows)
// Scan
area.GoTop()
for !area.EOF() {
// WHERE evaluation (if any). Fast variant — WHERE expressions
// compiled from SQL AST don't contain BEGIN SEQUENCE, so we can
// skip the defer/recover frame exit.
keep := true
if whereFn != nil {
hbrt.ExecPcodeFast(t, whereFn, nil)
keep = t.GetRetValue().AsBool()
}
if keep {
// Reserve nFields slots in flat, growing if needed.
off := len(flat)
end := off + nFields
if end > cap(flat) {
// Grow flat. Go's append growth policy handles this;
// we re-reserve space so the sub-slice math still holds.
flat = append(flat, make([]hbrt.Value, nFields)...)
} else {
flat = flat[:end]
}
row := flat[off:end:end]
// Collect column values directly into the backing buffer.
for i := 0; i < nFields; i++ {
// GetValue is 0-based
v, _ := area.GetValue(fieldPos[i] - 1)
row[i] = v
}
rows = append(rows, slab.WrapNext(row))
}
area.Skip(1)
}
t.PushValue(hbrt.MakeArrayFrom(rows))
t.RetValue()
}