PrecedenceParser

Defined in: packages/engine/src/parser/PrecedenceParser.ts:59

── Hybrid Precedence Climbing Parser ─────────────────────────────────────────

Two-tier dispatch strategy:

Tier 1 (Fast Path): Inline switch on token.typeId for built-in operators. - Prefix: NUMBER, IDENT, LPAREN, MINUS, PLUS, STRING - Infix: checked via static BP_TABLE (Uint8Array indexed by typeId) - No Map.get(), no parselet function call overhead for ~95% of tokens

Tier 2 (Plugin Path): ParseletRegistry fallback for custom/plugin tokens. - Prefix: Map.get(tokenTypeId) for PrefixParselet - Infix: Map.get(tokenTypeId) for InfixParselet - Full flexibility for custom syntax

The parser emits directly to a BytecodeBuilder, no AST intermediate. Implements the same public API as the legacy Parser class so existing parselets continue to work without modification.

Constructors

Constructor

new PrecedenceParser(
   parseletRegistry,
   maxDepth?,
   localeCode?,
   pluginFunctionIndex?,
   calendar?
): PrecedenceParser;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:172

Parameters

Parameter	Type	Default value
parseletRegistry	ParseletRegistry	undefined
maxDepth	number	50
localeCode	string	"en"
pluginFunctionIndex?	ReadonlyMap<string, number>	undefined
calendar?	CalendarBackend	DATE_CALENDAR

Returns

PrecedenceParser

Properties

infixMinBindingPower

infixMinBindingPower: number = 0;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:100

The binding power the current infix parselet is being invoked at, i.e. the minBp of the expression it sits inside. Set immediately before each Tier-2 parselet runs. A parselet that decides whether to swallow a following loose operator (UomLiteralParselet with a trailing in/to) reads this to respect precedence it cannot otherwise see: 120 km / 2 hours in kph must group as (120 km / 2 hours) in kph, not 120 km / (2 hours in kph).

---

BP_TABLE

readonly static BP_TABLE: Uint8Array;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:121

Static binding power table, built once at module load, shared across all instances. Index = tokenTypeId, value = binding power (0 = not a built-in infix).

Methods

consume()

consume(expectedType?): Token;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:1026

Parameters

Parameter	Type
expectedType?	string

Returns

Token

---

getCalendar()

getCalendar(): CalendarBackend;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:202

The calendar backend a parselet reads a date with at parse time: the engine’s own, so a fused literal reads back through the backend that built it. See the field’s doc for the sites that need it.

Returns

CalendarBackend

---

getLocaleCode()

getLocaleCode(): string;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:193

Get locale code for NumberParselet to normalize separators

Returns

string

---

load()

load(tokens, hasParens?): void;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:220

Load tokens for parsing. Identical to Parser.load().

Parameters

Parameter	Type	Description
tokens	Token[]	-
hasParens?	boolean	if false, skips the O(n) paren balance scan (~90% of expressions)

Returns

void

---

match()

match(expectedType): boolean;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:1052

Parameters

Parameter	Type
expectedType	string

Returns

boolean

---

parseExpression()

parseExpression(minBp?, _builder?): void;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:290

Parse a full expression starting at the current token position.

Parameters

Parameter	Type	Default value	Description
minBp	number	0	minimum binding power (precedence climbing threshold). For left-associative operators, the recursive call uses bp + 1. For right-associative operators (^), it uses bp - 1.
_builder?	BytecodeBuilder	undefined	accepted for parselet API compatibility; always uses this.builder.

Returns

void

---

peek()

peek(): Token | undefined;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:1061

Returns

Token | undefined

---

peekAt()

peekAt(offset): Token | undefined;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:1085

Read-only lookahead offset tokens past the current position, without consuming anything, peekAt(0) is equivalent to peek. this.tokens is a plain in-memory array (not a stream), so this is a simple, safe index read; no rewind/checkpoint mechanism is needed since nothing is consumed.

Parameters

Parameter	Type
offset	number

Returns

Token | undefined

---

previous()

previous(): Token | undefined;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:1089

Returns

Token | undefined

---

setBuilder()

setBuilder(builder): void;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:1102

Set the builder to use for the current parse. Called by ExpressionEngine.

Parameters

Parameter	Type
builder	BytecodeBuilder

Returns

void

---

setDiagnosticPipeline()

setDiagnosticPipeline(pipeline, expression): void;

Defined in: packages/engine/src/parser/PrecedenceParser.ts:210

Set the diagnostic pipeline for parselet matching events. Cleared after each non-cached parse to avoid holding refs.

Parameters

Parameter	Type
pipeline	DiagnosticPipeline | undefined
expression	string

Returns

void