com.nawforce.runtime.parsers.PageParser.scala Maven / Gradle / Ivy
/*
Copyright (c) 2019 Kevin Jones, All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
3. The name of the author may not be used to endorse or promote products
derived from this software without specific prior written permission.
*/
package com.nawforce.runtime.parsers
import com.nawforce.pkgforce.diagnostics.IssuesAnd
import com.nawforce.pkgforce.path.{PathLike, PathLocation}
import com.nawforce.runtime.parsers.PageParser.ParserRuleContext
import com.nawforce.runtime.parsers.antlr.{CharStreams, CommonTokenStream, ParseTree}
import com.nawforce.vfparser.{VFParser, VFLexer}
import scala.collection.compat.immutable.ArraySeq
import scala.reflect.ClassTag
import scala.scalajs.js
class PageParser(val source: Source) {
// We would like to extend this but it angers the JavaScript gods
private val is = source.asString
def parsePage(): IssuesAnd[VFParser.VfUnitContext] = {
parse(parser => parser.vfUnit())
}
/** Find a location for a rule, adapts based on source offsets to give absolute position in file
*/
def getPathLocation(context: ParserRuleContext): PathLocation = {
source.getLocation(context)
}
/** Extract the source used for a parser rule */
def extractSource(context: ParserRuleContext): Source = {
source.extractSource(context)
}
def parse[T](parse: VFParser => T): IssuesAnd[T] = {
val listener = new CollectingErrorListener(source.path)
val lexer = new VFLexer(CharStreams.fromString(is))
lexer.removeErrorListeners()
lexer.addErrorListener(listener)
val tokenStream = new CommonTokenStream(lexer)
tokenStream.fill()
val parser = new VFParser(tokenStream)
parser.removeErrorListeners()
parser.addErrorListener(listener)
val result = parse(parser)
IssuesAnd(listener.issues, result)
}
}
object PageParser {
type ParserRuleContext = com.nawforce.runtime.parsers.antlr.ParserRuleContext
type TerminalNode = com.nawforce.runtime.parsers.antlr.TerminalNode
def apply(path: PathLike, code: SourceData): PageParser = {
new PageParser(Source(path, code, 0, 0, None))
}
def clearCaches(): Unit = {
// Not supported
}
// Helper for JS Portability
def childCount(context: ParserRuleContext): Int = {
context.childCount
}
// Helper for JS Portability
def getText(context: ParseTree): String = {
context.text
}
// Helper for JS Portability
def getText(context: ParserRuleContext): String = {
if (context.childCount == 0) return ""
val builder = new StringBuilder
for (i <- 0 until context.childCount) {
builder.append(context.getChild(i).text)
}
builder.toString
}
// Helper for JS Portability
def getText(node: TerminalNode): String = {
node.text
}
// Helper for JS Portability
def toScala[T: ClassTag](collection: js.Array[T]): ArraySeq[T] = {
collection match {
case _ if collection.isEmpty => PageParser.emptyArraySeq
case _ => ArraySeq.unsafeWrapArray(collection.toArray)
}
}
// Helper for JS Portability
def toScala[T](value: js.UndefOr[T]): Option[T] = {
value.toOption
}
private val emptyArraySeq = ArraySeq()
}