simdb

A simple IMDB (in-memory database) generator for go.

We generate type-safe and thread-safe IMDB.

We generate table with columns and also implement some methods like Get, Insert, Update, Delete, etc...

Usage

Basic Informations.

Table Info

attribute	explanation	type
name	name of table	string	required
is_private	whether private or not	bool
is_pointer	whether the value is pointer or not	bool
key_type	type of key (this table's primary and an only unique key)	string	required
columns	array of columns	[]Columns

Column Info

attribute	explanation	type
name	name of column	string	required
type	type of column	string	requried
is_private	whether private or not	bool

Setting args

arg	explanation	positional or optional
file	generated file name	positional (1st arg)	default is `db.go`
dir	generated directory name	optional	default is current directory
package	package name of generated	optional	requried (default is `main`) [can set in config file]
config	config name withou extensions	optional	required (default is `simdb`)

Examples

foo/gen.go

package gen

//go:generate go run github.com/maru44/simdb db.go --dir=bar --package=bench --config=conf

foo/conf.yaml

name: bench
is_private: true
key_type: int
columns:
  - name: name
    type: string
  - name: email
    type: string
  - name: age
    type: uint
    is_private: true
  - name: is_vald
    type: bool

In this case, the generated file would be like following.

foo/bar/db.go

// Code generated by "github.com/maru44/simdb/gen"; DO NOT EDIT.

package bench

import (
	"fmt"
	"sync"
)

type (
	bench struct {
		Name    string
		Email   string
		age     uint
		IsValid bool
	}

	benchs struct {
		data map[int]bench
		sync.RWMutex
	}
)

func NewBenchs() benchs {
	return benchs{
		data: map[int]bench{},
	}
}

func (t *benchs) List() map[int]bench {
	t.RLock()
	defer t.RUnlock()
	return t.data
}

...

Performance

I measured bench mark. It depends on struct size, so for your information only.

Not pointer

I used this struct to measure bench mark.

type (
	bench struct {
		Name    string
		Email   string
		age     uint
		IsValid bool
	}

	benchs struct {
		data map[int]bench
		sync.RWMutex
	}
)

A little bit better than sync.Map.

goos: darwin
goarch: arm64
pkg: github.com/maru44/simdb/_tests/bench
BenchmarkInsert-8                4867156               235.6 ns/op           227 B/op          0 allocs/op
BenchmarkInsert_SyncMap-8        2842941               363.6 ns/op           184 B/op          5 allocs/op
BenchmarkGet-8                  12589046               113.1 ns/op             0 B/op          0 allocs/op
BenchmarkLoad-8                 12701738               111.3 ns/op             0 B/op          0 allocs/op
BenchmarkLoad_SyncMap-8          8800521               153.4 ns/op             0 B/op          0 allocs/op
BenchmarkDelete-8               12029200               120.3 ns/op             0 B/op          0 allocs/op
BenchmarkDelete_SyncMap-8       10508106               133.4 ns/op             0 B/op          0 allocs/op

Pointer

I used this struct to measure bench mark.

type (
	Pt struct {
		Name    string
		Email   string
		age     uint
		IsValid bool
	}

	Pts struct {
		data map[int]*Pt
		sync.RWMutex
	}
)

Even in the case of pointer, a little bit better than sync.Map.

goos: darwin
goarch: arm64
pkg: github.com/maru44/simdb/tests/bench
BenchmarkInsertPt-8             10145770               171.2 ns/op            67 B/op          0 allocs/op
BenchmarkInsertPt_SyncMap-8      3924448               388.6 ns/op           174 B/op          4 allocs/op
BenchmarkGetPt-8                23999539                71.32 ns/op            0 B/op          0 allocs/op
BenchmarkLoadPt-8               23471499                65.13 ns/op            0 B/op          0 allocs/op
BenchmarkLoadPt_SyncMap-8        9493228               153.6 ns/op             0 B/op          0 allocs/op
BenchmarkDeletePt-8             15360548                90.44 ns/op            0 B/op          0 allocs/op
BenchmarkDeletePt_SyncMap-8     10498572               133.5 ns/op             0 B/op          0 allocs/op

Comparing IsPointer or not

pointer

type (
	Pt struct {
		Name    string
		Email   string
		age     uint
		IsValid bool
	}

	Pts struct {
		data map[int]*Pt
		sync.RWMutex
	}
)

not pointer

type (
	bench struct {
		Name    string
		Email   string
		age     uint
		IsValid bool
	}

	benchs struct {
		data map[int]bench
		sync.RWMutex
	}
)

In this struct size, pointer is better than not pointer.

goos: darwin
goarch: arm64
pkg: github.com/maru44/simdb/tests/bench
BenchmarkInsert-8                4867156               235.6 ns/op           227 B/op          0 allocs/op
BenchmarkInsertPt-8             10145770               171.2 ns/op            67 B/op          0 allocs/op
BenchmarkGet-8                  12589046               113.1 ns/op             0 B/op          0 allocs/op
BenchmarkLoad-8                 12701738               111.3 ns/op             0 B/op          0 allocs/op
BenchmarkGetPt-8                23999539                71.32 ns/op            0 B/op          0 allocs/op
BenchmarkLoadPt-8               23471499                65.13 ns/op            0 B/op          0 allocs/op
BenchmarkDelete-8               12029200               120.3 ns/op             0 B/op          0 allocs/op
BenchmarkDeletePt-8             15360548                90.44 ns/op            0 B/op          0 allocs/op
BenchmarkUpdate-8                9903958               132.9 ns/op             0 B/op          0 allocs/op
BenchmarkUpdatePt-8             15739252                94.40 ns/op            0 B/op          0 allocs/op
BenchmarkUpdateName-8            7570064               175.3 ns/op             0 B/op          0 allocs/op
BenchmarkUpdateNamePt-8         18335049                82.61 ns/op            0 B/op          0 allocs/op

misinfo163/simdb