expr/parser.go

// Copyright (c) 2024-2026 Celestino Amoroso (celestino.amoroso@gmail.com).
// All rights reserved.

// parser.go
package expr

import (
	"errors"
	"fmt"
	"slices"

	"git.portale-stac.it/go-pkg/expr/scan"
	"golang.org/x/exp/constraints"
)

//-------- parser

type parserContext uint16

const (
	parserNoFlags                = 0
	allowMultiExpr parserContext = 1 << iota
	allowVarRef
	selectorContext
	listContext                                // squareContext for list
	indexContext                               // squareContext for index
	allowIndex                                 // allow index in squareContext
	squareContext = listContext | indexContext // Square parenthesis for list or index
)

func hasFlag[T constraints.Unsigned](set T, singleFlag T) bool {
	return (set & singleFlag) != 0
}

func addFlags[T constraints.Unsigned](set T, flags T) T {
	return set | flags
}

func addFlagsCond[T constraints.Unsigned](set T, flags T, cond bool) (newSet T) {
	if cond {
		newSet = set | flags
	} else {
		newSet = set
	}
	return
}

func remFlags[T constraints.Unsigned](set T, flags T) T {
	return set & (^flags)
}

type parser struct {
}

func NewParser() (p *parser) {
	p = &parser{}
	return p
}

func (parser *parser) Next(scanner *scan.Scanner) (tk *scan.Token) {
	for tk = scanner.Next(); tk.IsSymbol(scan.SymComment); tk = scanner.Next() {
	}
	return
}

func (parser *parser) parseFuncCall(scanner *scan.Scanner, ctx parserContext, tk *scan.Token) (tree *scan.Term, err error) {
	args := make([]*scan.Term, 0, 10)
	itemExpected := false
	lastSym := scan.SymUnknown
	for lastSym != scan.SymClosedRound && lastSym != scan.SymEos {
		var subTree *scan.Ast
		if subTree, err = parser.parseItem(scanner, ctx, scan.SymComma, scan.SymClosedRound); err != nil {
			break
		}
		prev := scanner.Previous()
		if subTree.Root() != nil {
			args = append(args, subTree.Root())
		} else if itemExpected {
			err = prev.ErrorExpectedGot("function-param-value")
			break
		}

		itemExpected = prev.Sym == scan.SymComma
		lastSym = scanner.Previous().Sym
	}
	if err == nil {
		if lastSym != scan.SymClosedRound {
			err = errors.New("unterminated arguments list")
		} else {
			tree = newFuncCallTerm(tk, args)
		}
	}
	return
}

func (parser *parser) parseFuncDef(scanner *scan.Scanner) (tree *scan.Term, err error) {
	// Example: "add = func(x,y) {x+y}
	var body *scan.Ast
	args := make([]*scan.Term, 0)
	lastSym := scan.SymUnknown
	defaultParamsStarted := false
	itemExpected := false
	tk := scanner.Previous()
	for lastSym != scan.SymClosedRound && lastSym != scan.SymEos {
		tk = parser.Next(scanner)
		if tk.IsSymbol(scan.SymIdentifier) {
			param := scan.NewTerm(tk)
			if len(args) > 0 {
				if pos := paramAlreadyDefined(args, param); pos > 0 {
					err = tk.Errorf("parameter %q at position %d already defined at position %d", param.Source(), len(args)+1, pos)
					break
				}
			}
			args = append(args, param)
			tk = parser.Next(scanner)
			if tk.Sym == scan.SymEqual {
				var paramExpr *scan.Ast
				defaultParamsStarted = true
				if paramExpr, err = parser.parseItem(scanner, parserNoFlags, scan.SymComma, scan.SymClosedRound); err != nil {
					break
				}
				param.ForceChild(paramExpr.Root())
			} else if defaultParamsStarted {
				err = tk.Errorf("can't mix default and non-default parameters")
				break
			}
		} else if itemExpected {
			prev := scanner.Previous()
			err = prev.ErrorExpectedGot("function-param-spec")
			break
		}
		lastSym = scanner.Previous().Sym
		itemExpected = lastSym == scan.SymComma
	}

	if err == nil && lastSym != scan.SymClosedRound {
		err = tk.ErrorExpectedGot(")")
	}
	if err == nil {
		tk = parser.Next(scanner)
		if tk.IsSymbol(scan.SymOpenBrace) {
			body, err = parser.parseGeneral(scanner, allowMultiExpr|allowVarRef, scan.SymClosedBrace)
		} else {
			err = tk.ErrorExpectedGot("{")
		}
	}
	if err == nil {
		if scanner.Previous().Sym != scan.SymClosedBrace {
			err = scanner.Previous().ErrorExpectedGot("}")
		} else {
			tk = scanner.MakeValueToken(scan.SymExpression, "", body)
			tree = newFuncDefTerm(tk, args)
		}
	}
	return
}

func paramAlreadyDefined(args []*scan.Term, param *scan.Term) (position int) {
	position = 0
	for i, arg := range args {
		if arg.Source() == param.Source() {
			position = i + 1
		}
	}
	return
}

func (parser *parser) parseList(scanner *scan.Scanner, ctx parserContext, termSym scan.Symbol) (listTerm *scan.Term, err error) {
	r, c := scanner.LastPos()
	args := make([]*scan.Term, 0)
	lastSym := scan.SymUnknown
	itemExpected := false
	itemCtx := remFlags(ctx, allowIndex)
	for lastSym != termSym && lastSym != scan.SymEos {
		zeroRequired := scanner.Current().Sym == scan.SymColon
		var itemTree *scan.Ast
		if itemTree, err = parser.parseItem(scanner, itemCtx, scan.SymComma, termSym); err == nil {
			root := itemTree.Root()
			if root != nil {
				if hasFlag(ctx, allowIndex) && root.Symbol() == scan.SymColon {
					changeColonToRange(root)
				}
				if !hasFlag(ctx, allowIndex) && root.Symbol() == scan.SymRange {
					// err = root.Errorf("unexpected range expression")
					err = errRangeUnexpectedExpression(root)
					break
				}
				args = append(args, root)
				if hasFlag(ctx, allowIndex) && root.Symbol() == scan.SymRange && zeroRequired { //len(root.children) == 0 {
					if len(root.Children) == 1 {
						root.Children = append(root.Children, root.Children[0])
					} else if len(root.Children) > 1 {
						// err = root.Errorf("invalid range specification")
						err = errRangeInvalidSpecification(root)
						break
					}
					zeroTk := scan.NewValueToken(root.Tk.Row(), root.Tk.Col(), scan.SymInteger, "0", int64(0))
					zeroTerm := scan.NewTerm(zeroTk)
					zeroTerm.SetParent(root)
					root.Children[0] = zeroTerm
				}
			} else if itemExpected {
				prev := scanner.Previous()
				err = prev.ErrorExpectedGot("list-item")
				break
			}
		} else {
			break
		}
		lastSym = scanner.Previous().Sym
		if itemExpected = lastSym == scan.SymComma; itemExpected {
			remFlags(ctx, allowIndex)
		}
	}
	if err == nil {
		if lastSym != termSym {
			err = scanner.Previous().ErrorExpectedGot("]")
		} else if termSym == scan.SymClosedSquare {
			listTerm = newListTerm(r, c, args)
		} else if termSym == scan.SymGreaterClosedSquare {
			listTerm = newLinkedListTerm(r, c, args)
		} else {
			err = fmt.Errorf("[%d:%d] unknown list type", r, c)
		}
	}
	return
}

func (parser *parser) parseIterDef(scanner *scan.Scanner, ctx parserContext) (subtree *scan.Term, err error) {
	tk := scanner.Previous()
	args := make([]*scan.Term, 0)
	lastSym := scan.SymUnknown
	itemExpected := false
	for lastSym != scan.SymClosedRound && lastSym != scan.SymEos {
		var subTree *scan.Ast
		if subTree, err = parser.parseItem(scanner, ctx, scan.SymComma, scan.SymClosedRound); err == nil {
			if subTree.Root() != nil {
				args = append(args, subTree.Root())
			} else if itemExpected {
				prev := scanner.Previous()
				err = prev.ErrorExpectedGot("iterator-param")
				break
			}
		} else {
			break
		}
		lastSym = scanner.Previous().Sym
		itemExpected = lastSym == scan.SymComma
	}
	if err == nil {
		if lastSym != scan.SymClosedRound {
			err = scanner.Previous().ErrorExpectedGot(")")
		} else {
			subtree = newIteratorTerm(tk, args)
		}
	}
	return
}

func (parser *parser) parseDictKey(scanner *scan.Scanner) (key any, err error) {
	tk := parser.Next(scanner)
	if tk.Sym == scan.SymError {
		err = tk.Error()
		return
	}
	if tk.Sym == scan.SymClosedBrace || tk.Sym == scan.SymEos {
		return
	}
	if tk.Sym == scan.SymInteger || tk.Sym == scan.SymString {
		tkSep := parser.Next(scanner)
		if tkSep.Sym != scan.SymColon {
			err = tkSep.ErrorExpectedGot(":")
		} else {
			key = tk.Value
		}
	} else {
		err = tk.ErrorExpectedGot("dictionary-key or }")
	}
	return
}

func (parser *parser) parseDictionary(scanner *scan.Scanner, ctx parserContext) (subtree *scan.Term, err error) {
	args := make(map[any]*scan.Term, 0)
	lastSym := scan.SymUnknown
	itemExpected := false
	for lastSym != scan.SymClosedBrace && lastSym != scan.SymEos {
		var subTree *scan.Ast
		var key any
		if key, err = parser.parseDictKey(scanner); err != nil {
			break
		} else if key == nil {
			tk := scanner.Previous()
			lastSym = tk.Sym
			if itemExpected {
				err = tk.ErrorExpectedGot("dictionary-key")
			}
			break
		}
		if subTree, err = parser.parseItem(scanner, ctx, scan.SymComma, scan.SymClosedBrace); err == nil {
			if subTree.Root() != nil {
				args[key] = subTree.Root()
			} else /*if key != nil*/ {
				prev := scanner.Previous()
				err = prev.ErrorExpectedGot("dictionary-value")
				break
			}
		} else {
			break
		}
		lastSym = scanner.Previous().Sym
		itemExpected = lastSym == scan.SymComma
	}
	if err == nil {
		if lastSym != scan.SymClosedBrace {
			err = scanner.Previous().ErrorExpectedGot("}")
		} else {
			subtree = newDictTerm(args)
		}
	}
	return
}

func (parser *parser) parseSelectorCase(scanner *scan.Scanner, ctx parserContext, defaultCase bool) (caseTerm *scan.Term, err error) {
	var filterList *scan.Term
	var caseExpr *scan.Ast
	ctx = remFlags(ctx, allowIndex)
	tk := parser.Next(scanner)
	startRow := tk.Row()
	startCol := tk.Col()
	if tk.Sym == scan.SymOpenSquare {
		if defaultCase {
			err = tk.Errorf("case list in default clause")
			return
		}
		if filterList, err = parser.parseList(scanner, remFlags(ctx, allowIndex), scan.SymClosedSquare); err != nil {
			return
		}
		tk = parser.Next(scanner)
		startRow = tk.Row()
		startCol = tk.Col()
	} else if !defaultCase {
		filterList = newListTerm(startRow, startCol, make([]*scan.Term, 0))
	}

	if tk.Sym == scan.SymOpenBrace {
		if caseExpr, err = parser.parseGeneral(scanner, ctx|allowMultiExpr, scan.SymClosedBrace); err != nil {
			return
		}
	} else {
		err = tk.ErrorExpectedGot("{")
	}

	if err == nil {
		caseTerm = newSelectorCaseTerm(startRow, startCol, filterList, caseExpr)
	}
	return
}

func addSelectorCase(selectorTerm, caseTerm *scan.Term) {
	if len(selectorTerm.Children) < 2 {
		caseListTerm := newListTermA(caseTerm)
		selectorTerm.Children = append(selectorTerm.Children, caseListTerm)
	} else {
		caseListTerm := selectorTerm.Children[1]
		caseList, _ := caseListTerm.Value().([]*scan.Term)
		caseList = append(caseList, caseTerm)
		caseListTerm.Tk.Value = caseList
	}
	caseTerm.Parent = selectorTerm
}

func (parser *parser) parseSelector(scanner *scan.Scanner, tree *scan.Ast, ctx parserContext) (selectorTerm *scan.Term, err error) {
	var caseTerm *scan.Term

	ctx = remFlags(ctx, allowIndex)
	tk := scanner.MakeToken(scan.SymSelector, '?')
	if selectorTerm, err = tree.AddToken(tk); err != nil {
		return
	}

	if caseTerm, err = parser.parseSelectorCase(scanner, ctx|allowVarRef, false); err == nil {
		addSelectorCase(selectorTerm, caseTerm)
	}
	return
}

func (parser *parser) parseItem(scanner *scan.Scanner, ctx parserContext, termSymbols ...scan.Symbol) (tree *scan.Ast, err error) {
	return parser.parseGeneral(scanner, ctx|allowVarRef, termSymbols...)
}

func (parser *parser) Parse(scanner *scan.Scanner, termSymbols ...scan.Symbol) (tree *scan.Ast, err error) {
	defer func() {
		if r := recover(); r != nil {
			if errVal, ok := r.(error); ok {
				err = errVal
			} else {
				err = errors.New("unexpected error while parsing the expression")
			}
		}
	}()
	termSymbols = append(termSymbols, scan.SymEos)
	return parser.parseGeneral(scanner, allowMultiExpr, termSymbols...)
}

func couldBeACollection(t *scan.Term) bool {
	var sym = scan.SymUnknown
	if t != nil {
		sym = t.Symbol()
	}
	// return sym == scan.SymList || sym == scan.SymString || sym == scan.SymDict || sym == scan.SymExpression || sym == scan.SymVariable
	return slices.Contains([]scan.Symbol{scan.SymList, scan.SymString, scan.SymDict, scan.SymExpression, scan.SymVariable, scan.SymIndex}, sym)
}

func listSubTree(tree *scan.Ast, listTerm *scan.Term, allowIndeces bool) (root *scan.Term, err error) {
	var tk *scan.Token
	if allowIndeces {
		tk = scan.NewToken(listTerm.Tk.Row(), listTerm.Tk.Col(), scan.SymIndex, listTerm.Source())
		root = scan.NewTerm(tk)
		if err = tree.AddTerm(root); err == nil {
			err = tree.AddTerm(listTerm)
		}
	} else {
		root = listTerm
		err = tree.AddTerm(listTerm)
	}
	return
}

func changePrefix(tk *scan.Token) {
	switch tk.Sym {
	case scan.SymMinus:
		tk.SetSymbol(scan.SymChangeSign)
	case scan.SymPlus:
		tk.SetSymbol(scan.SymUnchangeSign)
	case scan.SymStar:
		tk.SetSymbol(scan.SymDereference)
	case scan.SymExclamation:
		tk.SetSymbol(scan.SymNot)
	case scan.SymDoublePlus:
		tk.SetSymbol(scan.SymPreInc)
	case scan.SymDoubleMinus:
		tk.SetSymbol(scan.SymPreDec)
	}
}

func (parser *parser) parseGeneral(scanner *scan.Scanner, ctx parserContext, termSymbols ...scan.Symbol) (tree *scan.Ast, err error) {
	var selectorTerm *scan.Term = nil
	var currentTerm *scan.Term = nil
	var tk *scan.Token

	tree = scan.NewAst()
	firstToken := true
	// lastSym := SymUnknown
	for tk = parser.Next(scanner); err == nil && tk != nil && !tk.IsTerm(termSymbols); tk = parser.Next(scanner) {
		// if tk.Sym == SymComment {
		// 	continue
		// }

		if tk.Sym == scan.SymSemiColon {
			if hasFlag(ctx, allowMultiExpr) {
				tree.ToForest()
				firstToken = true
				currentTerm = nil
				selectorTerm = nil
				continue
			} else {
				err = tk.Errorf(`unexpected token %q, expected ",", "]", or ")"`, tk.Source())
				break
			}
		}

		//fmt.Println("Token:", tk)
		if firstToken {
			changePrefix(tk)
			firstToken = false
		}

		switch tk.Sym {
		case scan.SymOpenRound:
			var subTree *scan.Ast
			if subTree, err = parser.parseGeneral(scanner, ctx, scan.SymClosedRound); err == nil {
				if subTree.Root() == nil {
					err = tk.ErrorExpectedGotString("expression", "()")
				} else {
					exprTerm := newExprTerm(subTree.Root())
					err = tree.AddTerm(exprTerm)
					currentTerm = exprTerm
				}
			}
		case scan.SymFuncCall:
			var funcCallTerm *scan.Term
			if funcCallTerm, err = parser.parseFuncCall(scanner, ctx, tk); err == nil {
				err = tree.AddTerm(funcCallTerm)
				currentTerm = funcCallTerm
			}
		case scan.SymOpenSquare:
			var listTerm *scan.Term
			newCtx := addFlagsCond(addFlags(ctx, squareContext), allowIndex, couldBeACollection(currentTerm))
			if listTerm, err = parser.parseList(scanner, newCtx, scan.SymClosedSquare); err == nil {
				currentTerm, err = listSubTree(tree, listTerm, hasFlag(newCtx, allowIndex))
			}
		case scan.SymOpenSquareLess:
			var listTerm *scan.Term
			newCtx := addFlagsCond(addFlags(ctx, listContext), allowIndex, false)
			if listTerm, err = parser.parseList(scanner, newCtx, scan.SymGreaterClosedSquare); err == nil {
				currentTerm, err = listSubTree(tree, listTerm, hasFlag(newCtx, allowIndex))
			}
		case scan.SymOpenBrace:
			if currentTerm != nil && currentTerm.Symbol() == scan.SymColon {
				err = currentTerm.Errorf(`selector-case outside of a selector context`)
			} else {
				var mapTerm *scan.Term
				if mapTerm, err = parser.parseDictionary(scanner, ctx); err == nil {
					err = tree.AddTerm(mapTerm)
					currentTerm = mapTerm
				}
			}
		case scan.SymEqual, scan.SymPlusEqual, scan.SymMinusEqual, scan.SymStarEqual, scan.SymSlashEqual, scan.SymPercEqual, scan.SymAmpersandEqual, scan.SymVertBarEqual, scan.SymDoubleLessEqual, scan.SymDoubleGreaterEqual, scan.SymCaretEqual:
			currentTerm, err = tree.AddToken(tk)
			firstToken = true
		case scan.SymFuncDef:
			var funcDefTerm *scan.Term
			if funcDefTerm, err = parser.parseFuncDef(scanner); err == nil {
				err = tree.AddTerm(funcDefTerm)
				currentTerm = funcDefTerm
			}
		case scan.SymDollarRound:
			var iterDefTerm *scan.Term
			if iterDefTerm, err = parser.parseIterDef(scanner, ctx); err == nil {
				err = tree.AddTerm(iterDefTerm)
				currentTerm = iterDefTerm
			}
		case scan.SymIdentifier:
			if tk.Source()[0] == '@' && !hasFlag(ctx, allowVarRef) {
				err = tk.Errorf("variable references are not allowed in top level expressions: %q", tk.Source())
			} else {
				currentTerm, err = tree.AddToken(tk)
			}
		case scan.SymQuestion:
			if selectorTerm, err = parser.parseSelector(scanner, tree, ctx); err == nil {
				currentTerm = selectorTerm
				addFlags(ctx, selectorContext)
			}
		case scan.SymColon, scan.SymDoubleColon:
			var caseTerm *scan.Term
			if selectorTerm != nil {
				if caseTerm, err = parser.parseSelectorCase(scanner, ctx, tk.Sym == scan.SymDoubleColon); err == nil {
					addSelectorCase(selectorTerm, caseTerm)
					currentTerm = caseTerm
					if tk.Sym == scan.SymDoubleColon {
						selectorTerm = nil
					}
				}
			} else {
				currentTerm, err = tree.AddToken(tk)
				if tk.IsOneOfA(scan.SymColon, scan.SymRange) {
					// Colon outside a selector term acts like a separator
					firstToken = true
				}
			}
		default:
			currentTerm, err = tree.AddToken(tk)
		}

		if currentTerm != nil && currentTerm.Tk.Sym != scan.SymSelector && currentTerm.Parent != nil && currentTerm.Parent.Tk.Sym != scan.SymSelector {
			selectorTerm = nil
			remFlags(ctx, selectorContext)
		}
		// lastSym = tk.Sym
	}

	if err == nil {
		if !tk.IsOneOf(termSymbols) {
			var symDesc string
			if tk.IsSymbol(scan.SymError) {
				symDesc = tk.ErrorText()
			} else {
				symDesc = scan.SymToString(tk.Sym)
			}
			err = tk.ErrorExpectedGotStringWithPrefix("expected one of", scan.SymListToString(termSymbols, true), symDesc)
		} else {
			err = tk.Error()
		}
	}
	return
}