From c4f1d519ed7d816dd52303942b3e73a8e6f8e577 Mon Sep 17 00:00:00 2001
From: John Kerl <kerl.john.r@gmail.com>
Date: Sat, 26 Nov 2022 00:34:25 -0500
Subject: [PATCH] Use int8 for mvtype (memory reduction) (#1130)

* Use int8 for mvtype (memory reduction)

* Try to get around codespell false positives
---
 .github/workflows/codespell.yml      |  2 +-
 cmd/sizes/main.go                    | 19 +++++++++++++++++++
 internal/pkg/mlrval/mlrval_output.go | 17 +++++++++++++++++
 internal/pkg/mlrval/mlrval_type.go   | 13 ++++++-------
 4 files changed, 43 insertions(+), 8 deletions(-)
 create mode 100644 cmd/sizes/main.go

diff --git a/.github/workflows/codespell.yml b/.github/workflows/codespell.yml
index 60daa5045..497335281 100644
--- a/.github/workflows/codespell.yml
+++ b/.github/workflows/codespell.yml
@@ -33,4 +33,4 @@ jobs:
         with:
           check_filenames: true
           ignore_words_file: .codespellignore
-          skip: "*.csv,*.dkvp,*.txt,*.js,*.html,*.map,./tags,./test/cases,./docs/src/shapes-of-data.md.in,./docs/src/shapes-of-data.md"
+          skip: "*.csv,*.dkvp,*.txt,*.js,*.html,*.map,*.z,./tags,./test/cases,./docs/src/shapes-of-data.md.in,./docs/src/shapes-of-data.md,test/input/latin1.xtab"
diff --git a/cmd/sizes/main.go b/cmd/sizes/main.go
new file mode 100644
index 000000000..bddb09082
--- /dev/null
+++ b/cmd/sizes/main.go
@@ -0,0 +1,19 @@
+// ================================================================
+// Experiments for type-inference performance optimization
+// ================================================================
+
+// go build github.com/johnkerl/miller/cmd/sizes
+
+package main
+
+import (
+	"github.com/johnkerl/miller/internal/pkg/mlrval"
+)
+
+func main() {
+	var mvs [2]mlrval.Mlrval
+	mvs[0] = *mlrval.FromString("hello")
+	mvs[1] = *mlrval.FromString("world")
+	mvs[0].ShowSizes()
+	mvs[1].ShowSizes()
+}
diff --git a/internal/pkg/mlrval/mlrval_output.go b/internal/pkg/mlrval/mlrval_output.go
index 56912e5d9..4d6757340 100644
--- a/internal/pkg/mlrval/mlrval_output.go
+++ b/internal/pkg/mlrval/mlrval_output.go
@@ -3,6 +3,7 @@ package mlrval
 import (
 	"fmt"
 	"os"
+	"reflect"
 	"strconv"
 )
 
@@ -122,3 +123,19 @@ func (mv *Mlrval) StringifyValuesRecursively() {
 		mv.SetFromString(mv.String())
 	}
 }
+
+func (mv *Mlrval) ShowSizes() {
+	fmt.Printf("TOTAL            %p %d\n", mv, reflect.TypeOf(*mv).Size())
+	fmt.Printf("mv.intval        %p %d\n", &mv.intval, reflect.TypeOf(mv.intval).Size())
+	fmt.Printf("mv.floatval      %p %d\n", &mv.floatval, reflect.TypeOf(mv.floatval).Size())
+	fmt.Printf("mv.printrep      %p %d\n", &mv.printrep, reflect.TypeOf(mv.printrep).Size())
+	fmt.Printf("mv.printrepValid %p %d\n", &mv.printrepValid, reflect.TypeOf(mv.printrepValid).Size())
+	fmt.Printf("mv.boolval       %p %d\n", &mv.boolval, reflect.TypeOf(mv.boolval).Size())
+	fmt.Printf("mv.mvtype        %p %d\n", &mv.mvtype, reflect.TypeOf(mv.mvtype).Size())
+
+	fmt.Printf("mv.arrayval      %p %d\n", &mv.arrayval, reflect.TypeOf(mv.arrayval).Size())
+	fmt.Printf("mv.mapval        %p %d\n", &mv.mapval, reflect.TypeOf(mv.mapval).Size())
+	if mv.funcval != nil {
+		fmt.Printf("mv.funcval       %p %d\n", &mv.funcval, reflect.TypeOf(mv.funcval).Size())
+	}
+}
diff --git a/internal/pkg/mlrval/mlrval_type.go b/internal/pkg/mlrval/mlrval_type.go
index 9e38ead8f..a752360a0 100644
--- a/internal/pkg/mlrval/mlrval_type.go
+++ b/internal/pkg/mlrval/mlrval_type.go
@@ -54,17 +54,16 @@
 package mlrval
 
 type Mlrval struct {
+	intval        int64
+	floatval      float64
+	printrep      string
+	printrepValid bool
+	boolval       bool
 
 	// Enumeration for string / int / float / boolean / etc.
 	// I would call this "type" not "mvtype" but "type" is a keyword in Go.
 	mvtype MVType
 
-	printrep      string
-	printrepValid bool
-	intval        int64
-	floatval      float64
-	boolval       bool
-
 	arrayval []*Mlrval
 	mapval   *Mlrmap
 	// First-class-function literals from internal/pkg/dsl/cst.
@@ -84,7 +83,7 @@ const ABSENT_PRINTREP = "(absent)"
 // JavaScript, similarly, there are undefined and null, respectively --
 // Miller's absent is more like JavaScript's undefined.)
 
-type MVType int
+type MVType int8
 
 // Important: the values of these enums are used to index into disposition
 // matrices. If they are changed, it will break the disposition matrices, or