mirror of
https://codeberg.org/forgejo/forgejo.git
synced 2024-11-24 08:57:03 -05:00
5b89670a31
One of the proposals in #23328 This PR introduces a simple expression calculator (templates/eval/eval.go), it can do basic expression calculations. Many untested template helper functions like `Mul` `Add` can be replaced by this new approach. Then these `Add` / `Mul` / `percentage` / `Subtract` / `DiffStatsWidth` could all use this `Eval`. And it provides enhancements for Golang templates, and improves readability. Some examples: ---- * Before: `{{Add (Mul $glyph.Row 12) 12}}` * After: `{{Eval $glyph.Row "*" 12 "+" 12}}` ---- * Before: `{{if lt (Add $i 1) (len $.Topics)}}` * After: `{{if Eval $i "+" 1 "<" (len $.Topics)}}` ## FAQ ### Why not use an existing expression package? We need a highly customized expression engine: * do the calculation on the fly, without pre-compiling * deal with int/int64/float64 types, to make the result could be used in Golang template. * make the syntax could be used in the Golang template directly * do not introduce too much complex or strange syntax, we just need a simple calculator. * it needs to strictly follow Golang template's behavior, for example, Golang template treats all non-zero values as truth, but many 3rd packages don't do so. ### What's the benefit? * Developers don't need to add more `Add`/`Mul`/`Sub`-like functions, they were getting more and more. Now, only one `Eval` is enough for all cases. * The new code reads better than old `{{Add (Mul $glyph.Row 12) 12}}`, the old one isn't familiar to most procedural programming developers (eg, the Golang expression syntax). * The `Eval` is fully covered by tests, many old `Add`/`Mul`-like functions were never tested. ### The performance? It doesn't use `reflect`, it doesn't need to parse or compile when used in Golang template, the performance is as fast as native Go template. ### Is it too complex? Could it be unstable? The expression calculator program is a common homework for computer science students, and it's widely used as a teaching and practicing purpose for developers. The algorithm is pretty well-known. The behavior can be clearly defined, it is stable.
344 lines
7.5 KiB
Go
344 lines
7.5 KiB
Go
// Copyright 2023 The Gitea Authors. All rights reserved.
|
|
// SPDX-License-Identifier: MIT
|
|
|
|
package eval
|
|
|
|
import (
|
|
"fmt"
|
|
"strconv"
|
|
"strings"
|
|
|
|
"code.gitea.io/gitea/modules/util"
|
|
)
|
|
|
|
type Num struct {
|
|
Value any // int64 or float64, nil on error
|
|
}
|
|
|
|
var opPrecedence = map[string]int{
|
|
// "(": 1, this is for low precedence like function calls, they are handled separately
|
|
"or": 2,
|
|
"and": 3,
|
|
"not": 4,
|
|
"==": 5, "!=": 5, "<": 5, "<=": 5, ">": 5, ">=": 5,
|
|
"+": 6, "-": 6,
|
|
"*": 7, "/": 7,
|
|
}
|
|
|
|
type stack[T any] struct {
|
|
name string
|
|
elems []T
|
|
}
|
|
|
|
func (s *stack[T]) push(t T) {
|
|
s.elems = append(s.elems, t)
|
|
}
|
|
|
|
func (s *stack[T]) pop() T {
|
|
if len(s.elems) == 0 {
|
|
panic(s.name + " stack is empty")
|
|
}
|
|
t := s.elems[len(s.elems)-1]
|
|
s.elems = s.elems[:len(s.elems)-1]
|
|
return t
|
|
}
|
|
|
|
func (s *stack[T]) peek() T {
|
|
if len(s.elems) == 0 {
|
|
panic(s.name + " stack is empty")
|
|
}
|
|
return s.elems[len(s.elems)-1]
|
|
}
|
|
|
|
type operator string
|
|
|
|
type eval struct {
|
|
stackNum stack[Num]
|
|
stackOp stack[operator]
|
|
funcMap map[string]func([]Num) Num
|
|
}
|
|
|
|
func newEval() *eval {
|
|
e := &eval{}
|
|
e.stackNum.name = "num"
|
|
e.stackOp.name = "op"
|
|
return e
|
|
}
|
|
|
|
func toNum(v any) (Num, error) {
|
|
switch v := v.(type) {
|
|
case string:
|
|
if strings.Contains(v, ".") {
|
|
n, err := strconv.ParseFloat(v, 64)
|
|
if err != nil {
|
|
return Num{n}, err
|
|
}
|
|
return Num{n}, nil
|
|
}
|
|
n, err := strconv.ParseInt(v, 10, 64)
|
|
if err != nil {
|
|
return Num{n}, err
|
|
}
|
|
return Num{n}, nil
|
|
case float32, float64:
|
|
n, _ := util.ToFloat64(v)
|
|
return Num{n}, nil
|
|
default:
|
|
n, err := util.ToInt64(v)
|
|
if err != nil {
|
|
return Num{n}, err
|
|
}
|
|
return Num{n}, nil
|
|
}
|
|
}
|
|
|
|
func truth(b bool) int64 {
|
|
if b {
|
|
return int64(1)
|
|
}
|
|
return int64(0)
|
|
}
|
|
|
|
func applyOp2Generic[T int64 | float64](op operator, n1, n2 T) Num {
|
|
switch op {
|
|
case "+":
|
|
return Num{n1 + n2}
|
|
case "-":
|
|
return Num{n1 - n2}
|
|
case "*":
|
|
return Num{n1 * n2}
|
|
case "/":
|
|
return Num{n1 / n2}
|
|
case "==":
|
|
return Num{truth(n1 == n2)}
|
|
case "!=":
|
|
return Num{truth(n1 != n2)}
|
|
case "<":
|
|
return Num{truth(n1 < n2)}
|
|
case "<=":
|
|
return Num{truth(n1 <= n2)}
|
|
case ">":
|
|
return Num{truth(n1 > n2)}
|
|
case ">=":
|
|
return Num{truth(n1 >= n2)}
|
|
case "and":
|
|
t1, _ := util.ToFloat64(n1)
|
|
t2, _ := util.ToFloat64(n2)
|
|
return Num{truth(t1 != 0 && t2 != 0)}
|
|
case "or":
|
|
t1, _ := util.ToFloat64(n1)
|
|
t2, _ := util.ToFloat64(n2)
|
|
return Num{truth(t1 != 0 || t2 != 0)}
|
|
}
|
|
panic("unknown operator: " + string(op))
|
|
}
|
|
|
|
func applyOp2(op operator, n1, n2 Num) Num {
|
|
float := false
|
|
if _, ok := n1.Value.(float64); ok {
|
|
float = true
|
|
} else if _, ok = n2.Value.(float64); ok {
|
|
float = true
|
|
}
|
|
if float {
|
|
f1, _ := util.ToFloat64(n1.Value)
|
|
f2, _ := util.ToFloat64(n2.Value)
|
|
return applyOp2Generic(op, f1, f2)
|
|
}
|
|
return applyOp2Generic(op, n1.Value.(int64), n2.Value.(int64))
|
|
}
|
|
|
|
func toOp(v any) (operator, error) {
|
|
if v, ok := v.(string); ok {
|
|
return operator(v), nil
|
|
}
|
|
return "", fmt.Errorf(`unsupported token type "%T"`, v)
|
|
}
|
|
|
|
func (op operator) hasOpenBracket() bool {
|
|
return strings.HasSuffix(string(op), "(") // it's used to support functions like "sum("
|
|
}
|
|
|
|
func (op operator) isComma() bool {
|
|
return op == ","
|
|
}
|
|
|
|
func (op operator) isCloseBracket() bool {
|
|
return op == ")"
|
|
}
|
|
|
|
type ExprError struct {
|
|
msg string
|
|
tokens []any
|
|
err error
|
|
}
|
|
|
|
func (err ExprError) Error() string {
|
|
sb := strings.Builder{}
|
|
sb.WriteString(err.msg)
|
|
sb.WriteString(" [ ")
|
|
for _, token := range err.tokens {
|
|
_, _ = fmt.Fprintf(&sb, `"%v" `, token)
|
|
}
|
|
sb.WriteString("]")
|
|
if err.err != nil {
|
|
sb.WriteString(": ")
|
|
sb.WriteString(err.err.Error())
|
|
}
|
|
return sb.String()
|
|
}
|
|
|
|
func (err ExprError) Unwrap() error {
|
|
return err.err
|
|
}
|
|
|
|
func (e *eval) applyOp() {
|
|
op := e.stackOp.pop()
|
|
if op == "not" {
|
|
num := e.stackNum.pop()
|
|
i, _ := util.ToInt64(num.Value)
|
|
e.stackNum.push(Num{truth(i == 0)})
|
|
} else if op.hasOpenBracket() || op.isCloseBracket() || op.isComma() {
|
|
panic(fmt.Sprintf("incomplete sub-expression with operator %q", op))
|
|
} else {
|
|
num2 := e.stackNum.pop()
|
|
num1 := e.stackNum.pop()
|
|
e.stackNum.push(applyOp2(op, num1, num2))
|
|
}
|
|
}
|
|
|
|
func (e *eval) exec(tokens ...any) (ret Num, err error) {
|
|
defer func() {
|
|
if r := recover(); r != nil {
|
|
rErr, ok := r.(error)
|
|
if !ok {
|
|
rErr = fmt.Errorf("%v", r)
|
|
}
|
|
err = ExprError{"invalid expression", tokens, rErr}
|
|
}
|
|
}()
|
|
for _, token := range tokens {
|
|
n, err := toNum(token)
|
|
if err == nil {
|
|
e.stackNum.push(n)
|
|
continue
|
|
}
|
|
|
|
op, err := toOp(token)
|
|
if err != nil {
|
|
return Num{}, ExprError{"invalid expression", tokens, err}
|
|
}
|
|
|
|
switch {
|
|
case op.hasOpenBracket():
|
|
e.stackOp.push(op)
|
|
case op.isCloseBracket(), op.isComma():
|
|
var stackTopOp operator
|
|
for len(e.stackOp.elems) > 0 {
|
|
stackTopOp = e.stackOp.peek()
|
|
if stackTopOp.hasOpenBracket() || stackTopOp.isComma() {
|
|
break
|
|
}
|
|
e.applyOp()
|
|
}
|
|
if op.isCloseBracket() {
|
|
nums := []Num{e.stackNum.pop()}
|
|
for !e.stackOp.peek().hasOpenBracket() {
|
|
stackTopOp = e.stackOp.pop()
|
|
if !stackTopOp.isComma() {
|
|
return Num{}, ExprError{"bracket doesn't match", tokens, nil}
|
|
}
|
|
nums = append(nums, e.stackNum.pop())
|
|
}
|
|
for i, j := 0, len(nums)-1; i < j; i, j = i+1, j-1 {
|
|
nums[i], nums[j] = nums[j], nums[i] // reverse nums slice, to get the right order for arguments
|
|
}
|
|
stackTopOp = e.stackOp.pop()
|
|
fn := string(stackTopOp[:len(stackTopOp)-1])
|
|
if fn == "" {
|
|
if len(nums) != 1 {
|
|
return Num{}, ExprError{"too many values in one bracket", tokens, nil}
|
|
}
|
|
e.stackNum.push(nums[0])
|
|
} else if f, ok := e.funcMap[fn]; ok {
|
|
e.stackNum.push(f(nums))
|
|
} else {
|
|
return Num{}, ExprError{"unknown function: " + fn, tokens, nil}
|
|
}
|
|
} else {
|
|
e.stackOp.push(op)
|
|
}
|
|
default:
|
|
for len(e.stackOp.elems) > 0 && len(e.stackNum.elems) > 0 {
|
|
stackTopOp := e.stackOp.peek()
|
|
if stackTopOp.hasOpenBracket() || stackTopOp.isComma() || precedence(stackTopOp, op) < 0 {
|
|
break
|
|
}
|
|
e.applyOp()
|
|
}
|
|
e.stackOp.push(op)
|
|
}
|
|
}
|
|
for len(e.stackOp.elems) > 0 && !e.stackOp.peek().isComma() {
|
|
e.applyOp()
|
|
}
|
|
if len(e.stackNum.elems) != 1 {
|
|
return Num{}, ExprError{fmt.Sprintf("expect 1 value as final result, but there are %d", len(e.stackNum.elems)), tokens, nil}
|
|
}
|
|
return e.stackNum.pop(), nil
|
|
}
|
|
|
|
func precedence(op1, op2 operator) int {
|
|
p1 := opPrecedence[string(op1)]
|
|
p2 := opPrecedence[string(op2)]
|
|
if p1 == 0 {
|
|
panic("unknown operator precedence: " + string(op1))
|
|
} else if p2 == 0 {
|
|
panic("unknown operator precedence: " + string(op2))
|
|
}
|
|
return p1 - p2
|
|
}
|
|
|
|
func castFloat64(nums []Num) bool {
|
|
hasFloat := false
|
|
for _, num := range nums {
|
|
if _, hasFloat = num.Value.(float64); hasFloat {
|
|
break
|
|
}
|
|
}
|
|
if hasFloat {
|
|
for i, num := range nums {
|
|
if _, ok := num.Value.(float64); !ok {
|
|
f, _ := util.ToFloat64(num.Value)
|
|
nums[i] = Num{f}
|
|
}
|
|
}
|
|
}
|
|
return hasFloat
|
|
}
|
|
|
|
func fnSum(nums []Num) Num {
|
|
if castFloat64(nums) {
|
|
var sum float64
|
|
for _, num := range nums {
|
|
sum += num.Value.(float64)
|
|
}
|
|
return Num{sum}
|
|
}
|
|
var sum int64
|
|
for _, num := range nums {
|
|
sum += num.Value.(int64)
|
|
}
|
|
return Num{sum}
|
|
}
|
|
|
|
// Expr evaluates the given expression tokens and returns the result.
|
|
// It supports the following operators: +, -, *, /, and, or, not, ==, !=, >, >=, <, <=.
|
|
// Non-zero values are treated as true, zero values are treated as false.
|
|
// If no error occurs, the result is either an int64 or a float64.
|
|
// If all numbers are integer, the result is an int64, otherwise if there is any float number, the result is a float64.
|
|
func Expr(tokens ...any) (Num, error) {
|
|
e := newEval()
|
|
e.funcMap = map[string]func([]Num) Num{"sum": fnSum}
|
|
return e.exec(tokens...)
|
|
}
|