package scalatex
package stages
import org.parboiled2._
import torimatomeru.ScalaSyntax
import Util._
/**
* Parses the input text into a roughly-structured AST. This AST
* is much simpler than the real Scala AST, but serves us well
* enough until we stuff the code-strings into the real Scala
* parser later
*/
object Parser extends (String => Ast.Block){
def apply(input: String): Ast.Block = {
new Parser(input).Body.run().get
}
}
class Parser(input: ParserInput, indent: Int = 0) extends ScalaSyntax(input) {
val txt = input.sliceString(0, input.length)
val indentTable = txt.split('\n').map{ s =>
if (s.trim == "") -1
else s.takeWhile(_ == ' ').length
}
val nextIndentTable = (0 until indentTable.length).map { i =>
val index = indentTable.indexWhere(_ != -1, i + 1)
if (index == -1) 100000
else indentTable(index)
}
def cursorNextIndent() = {
nextIndentTable(txt.take(cursor).count(_ == '\n'))
}
def TextNot(chars: String) = rule {
capture(oneOrMore(noneOf(chars + "\n") | "@@")) ~> {
x => Ast.Block.Text(x.replace("@@", "@"))
}
}
def Text = TextNot("@")
def Code = rule {
"@" ~ capture(Id | BlockExpr2 | ('(' ~ optional(Exprs) ~ ')'))
}
def Header = rule {
"@" ~ capture(Def | Import)
}
def HeaderBlock: Rule1[Ast.Block] = rule{
oneOrMore(Header ~ NewlineS) ~ runSubParser{new Parser(_, indent).Body0} ~> {
(head: Seq[String], body: Ast.Block) => body.copy(front = Some(head.mkString("\n")))
}
}
def BlankLine = rule{ '\n' ~ zeroOrMore(' ') ~ &('\n') }
def Indent = rule{ '\n' ~ indent.times(' ') ~ zeroOrMore(' ') }
def LoneScalaChain: Rule2[Ast.Block.Text, Ast.Chain] = rule {
(capture(Indent) ~> (Ast.Block.Text(_))) ~
ScalaChain ~
zeroOrMore(BlankLine) ~
test(cursorNextIndent() > indent) ~
runSubParser {
new Parser(_, cursorNextIndent()).Body
} ~> { (chain: Ast.Chain, body: Ast.Block) =>
chain.copy(parts = chain.parts :+ body)
}
}
def ScalaChain = rule {
Code ~ zeroOrMore(Extension) ~> {Ast.Chain(_, _)}
}
def Extension: Rule1[Ast.Chain.Sub] = rule {
('.' ~ capture(Id) ~> (Ast.Chain.Prop(_))) |
(capture(TypeArgs2) ~> (Ast.Chain.TypeArgs(_))) |
(capture(ArgumentExprs2) ~> (Ast.Chain.Args(_))) |
TBlock
}
def Ws = Whitespace
// clones of the version in ScalaSyntax, but without tailing whitespace or newlines
def TypeArgs2 = rule { '[' ~ Ws ~ Types ~ ']' }
def ArgumentExprs2 = rule {
'(' ~ Ws ~ (optional(Exprs ~ ',' ~ Ws) ~ PostfixExpr ~ ':' ~ Ws ~ '_' ~ Ws ~ '*' ~ Ws | optional(Exprs)) ~ ')'
}
def BlockExpr2: Rule0 = rule { '{' ~ Ws ~ (CaseClauses | Block) ~ '}' }
def TBlock = rule{ '{' ~ Body ~ '}' }
def BodyItem = rule{
LoneScalaChain ~> (Seq(_, _)) |
HeaderBlock ~> (Seq(_)) |
TextNot("@}") ~> (Seq(_)) |
(capture(Indent) ~> (x => Seq(Ast.Block.Text(x)))) |
(capture(BlankLine) ~> (x => Seq(Ast.Block.Text(x)))) |
ScalaChain ~> (Seq(_))
}
def Body = rule{
oneOrMore(BodyItem) ~> {x =>
Ast.Block(None, x.flatten)
}
}
def Body0 = rule{
zeroOrMore(BodyItem) ~> {x =>
Ast.Block(None, x.flatten)
}
}
}
trait Ast{
def offset: Int
}
object Ast{
/**
* @param front Any parameter lists (if it's a lambda), imports, val/def/lazy-val,
* class/object/trait declarations that occur before the parts of this
* block, and must be in scope within those parts
* @param parts The various bits of text and other things which make up this block
* @param offset
*/
case class Block(front: Option[String],
parts: Seq[Block.Sub],
offset: Int = 0)
extends Chain.Sub with Block.Sub
object Block{
trait Sub extends Ast
case class Text(txt: String, offset: Int = 0) extends Block.Sub
}
/**
* @param lhs The first expression in this method-chain
* @param parts A list of follow-on items chained to the first
* @param offset
*/
case class Chain(lhs: String, parts: Seq[Chain.Sub], offset: Int = 0) extends Block.Sub
object Chain{
trait Sub extends Ast
case class Prop(str: String, offset: Int = 0) extends Sub
case class TypeArgs(str: String, offset: Int = 0) extends Sub
case class Args(str: String, offset: Int = 0) extends Sub
}
}