refactor(LNG-147): Rewrite list to tree conversion tail recusive (#711)

* Implemented ListToTreeConverter
This commit is contained in:
InversionSpaces 2023-05-16 12:17:30 +02:00 committed by GitHub
parent 440899f6c6
commit 3af28d5e53
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 198 additions and 132 deletions

View File

@ -15,6 +15,8 @@ import cats.parse.{Parser as P, Parser0 as P0}
import cats.syntax.comonad.* import cats.syntax.comonad.*
import cats.{~>, Comonad, Eval} import cats.{~>, Comonad, Eval}
import scribe.Logging import scribe.Logging
import aqua.parser.Ast.Tree
import aqua.parser.ListToTreeConverter
abstract class Expr[F[_]](val companion: Expr.Companion, val token: Token[F]) { abstract class Expr[F[_]](val companion: Expr.Companion, val token: Token[F]) {
@ -92,143 +94,19 @@ object Expr {
abstract class AndIndented extends Block with Logging { abstract class AndIndented extends Block with Logging {
def validChildren: List[Lexem] def validChildren: List[Lexem]
private def leaf[F[_]](expr: Expr[F]): Ast.Tree[F] =
Cofree[Chain, Expr[F]](
expr,
Eval.now(Chain.empty)
)
private def last[F[_]](tree: Ast.Tree[F]): Expr[F] =
tree.tailForced.lastOption.fold(tree.head)(last)
private def setLeafs[F[_]](tree: Ast.Tree[F], children: Chain[Tree[F]]): Tree[F] =
tree.copy(tail = tree.tail.map {
case pref :== last =>
pref :+ setLeafs(last, children)
case _ =>
children
})
// Check if expression can be added in current block
private def canAddToBlock[F[_]](block: Tree[F], expr: Expr[F]): Boolean = {
block.head.companion match {
case b: AndIndented =>
b.validChildren.map {
case ll: LazyLexem => ll.c
case vc => vc
}.contains(expr.companion)
case _: Prefix =>
block.tail.value.headOption.exists(t => canAddToBlock(t, expr))
case _ => false
}
}
// Generate error if expression (child) cannot be added to a block
private def wrongChildError[F[_]](indent: F[String], expr: Expr[F]): ParserError[F] = {
val msg = expr match {
case ReturnExpr(_) =>
"Return expression must be on the top indentation level and at the end of function body"
// could there be other expressions?
case _ => "This expression is on the wrong indentation level"
}
BlockIndentError(indent, msg)
}
private def headIsBlock[F[_]](tree: Tree[F]): Boolean = {
tree.tail.value.headOption match {
case Some(t) => t.head.isBlock
case _ => tree.head.isBlock
}
}
private case class Acc[F[_]](
block: Tree[F],
initialIndentF: F[String],
tail: Chain[(F[String], Ast.Tree[F])] = Chain.empty[(F[String], Ast.Tree[F])],
window: Chain[Tree[F]] = Chain.empty[Tree[F]],
errors: Chain[ParserError[F]] = Chain.empty[ParserError[F]]
)
// converts list of expressions to a tree of tokens
private def listToTree[F[_]: Comonad: LiftParser](
acc: Acc[F]
): ValidatedNec[ParserError[F], Acc[F]] = {
val initialIndent = acc.initialIndentF.extract.length
acc.tail.uncons match {
case Some(((currentIndent, currentExpr), tail)) =>
val current = last(currentExpr)
// if current indent is bigger then block indentation
// then add current expression to this block
if (currentIndent.extract.length > initialIndent) {
// if current expression is a block, create tree of this block and return remaining tail
if (headIsBlock(currentExpr)) {
listToTree(Acc(currentExpr, currentIndent, tail, errors = acc.errors)).andThen {
case a@Acc(innerTree, _, newTail, window, errors) =>
if (window.nonEmpty) {
logger.warn("Internal: Window cannot be empty after converting list of expressions to a tree.")
logger.warn("Current state: " + a)
}
listToTree(
acc.copy(
window = acc.window :+ innerTree,
tail = newTail,
errors = acc.errors ++ errors
)
)
}
} else {
// if expression not a block, add it to a window until we meet the end of the block
if (canAddToBlock(acc.block, current)) {
listToTree(acc.copy(window = acc.window :+ currentExpr, tail = tail))
} else {
val error = wrongChildError(currentIndent, current)
validNec(acc.copy(tail = tail, errors = acc.errors :+ error))
}
}
} else {
val errors = if (acc.window.isEmpty) {
// error if a block is empty
val error = BlockIndentError(acc.initialIndentF, "Block expression has no body")
acc.errors :+ error
} else acc.errors
// if current indentation less or equal to block indentation,
// add all expressions in window to a head
validNec(
Acc(
setLeafs(acc.block, acc.window),
acc.initialIndentF,
(currentIndent, currentExpr) +: tail,
errors = errors
)
)
}
case None =>
// end of top-level block
NonEmptyChain
.fromChain(acc.errors)
.map(invalid)
.getOrElse(validNec(Acc(setLeafs(acc.block, acc.window), acc.initialIndentF)))
}
}
override lazy val ast: P[ValidatedNec[ParserError[Span.S], Ast.Tree[Span.S]]] = override lazy val ast: P[ValidatedNec[ParserError[Span.S], Ast.Tree[Span.S]]] =
(readLine ~ (` \n+` *> (readLine ~ (` \n+` *>
(P.repSep( (P.repSep(
` `.lift ~ P.oneOf(validChildren.map(_.readLine.backtrack)), ` `.lift ~ P.oneOf(validChildren.map(_.readLine.backtrack)),
` \n+` ` \n+`
) <* ` \n`.?))).map { t => ) <* ` \n`.?))).map { case (open, lines) =>
val startIndent = t._1.head.token.as("") lines
listToTree(Acc(t._1, startIndent, Chain.fromSeq(t._2.toList))).map { res => .foldLeft(
res._1 ListToTreeConverter(open)
} ) { case (acc, (indent, line)) =>
acc.next(indent, line)
}
.result
} }
} }
} }

View File

@ -0,0 +1,188 @@
package aqua.parser
import aqua.parser.Ast.Tree
import aqua.parser.Expr.{AndIndented, LazyLexem, Prefix}
import aqua.parser.expr.func.ReturnExpr
import aqua.parser.lift.LiftParser
import cats.Comonad
import cats.data.Chain
import cats.data.Chain.:==
import cats.syntax.comonad.*
import cats.data.{NonEmptyChain, ValidatedNec}
import cats.data.Validated.*
import aqua.parser.Expr.LazyLexem
/**
* Converts a list of lines to a tree
*/
final case class ListToTreeConverter[F[_]](
currentBlock: ListToTreeConverter.Block[F], // Current block data
stack: List[ListToTreeConverter.Block[F]] = Nil, // Stack of opened blocks
errors: Chain[ParserError[F]] = Chain.empty[ParserError[F]] // Errors
)(using Comonad[F]) {
// Import helper functions
import ListToTreeConverter.*
private def addError(error: ParserError[F]): ListToTreeConverter[F] =
copy(errors = errors :+ error)
private def pushBlock(indent: F[String], line: Tree[F]): ListToTreeConverter[F] =
copy(currentBlock = Block(indent, line), stack = currentBlock :: stack)
private def addToCurrentBlock(line: Tree[F]): ListToTreeConverter[F] =
copy(currentBlock = currentBlock.add(line))
private def popBlock: Option[ListToTreeConverter[F]] =
stack match {
case Nil => None
case prevBlock :: tail =>
Some(copy(currentBlock = prevBlock.add(currentBlock.close), stack = tail))
}
/**
* Method to call on each new line
*/
@scala.annotation.tailrec
def next(indent: F[String], line: Tree[F]): ListToTreeConverter[F] =
if (indentValue(indent) > indentValue(currentBlock.indent)) {
if (isBlock(line)) {
pushBlock(indent, line)
} else {
val expr = lastExpr(line)
if (currentBlock.canAdd(expr)) {
addToCurrentBlock(line)
} else {
addError(wrongChildError(indent, expr))
}
}
} else {
val emptyChecked = if (currentBlock.isEmpty) {
addError(emptyBlockError(currentBlock.indent))
} else this
emptyChecked.popBlock match {
case Some(blockPopped) => blockPopped.next(indent, line)
// This should not happen because of the way of parsing
case _ => emptyChecked.addError(unexpectedIndentError(indent))
}
}
/**
* Produce the result of the conversion
*/
@scala.annotation.tailrec
def result: ValidatedNec[ParserError[F], Tree[F]] =
popBlock match {
case Some(blockPopped) => blockPopped.result
case _ =>
NonEmptyChain
.fromChain(errors)
.map(invalid)
.getOrElse(
valid(currentBlock.close)
)
}
}
object ListToTreeConverter {
/**
* Constructs a converter from block opening line
*/
def apply[F[_]](open: Tree[F])(using Comonad[F]): ListToTreeConverter[F] =
ListToTreeConverter(Block(open.head.token.as(""), open))
/**
* Data associated with a block
*/
final case class Block[F[_]](
indent: F[String], // Indentation of the block opening line
block: Tree[F], // Block opening line
content: Chain[Tree[F]] = Chain.empty[Tree[F]] // Children of the block
) {
/**
* Check if expr can be added to this block
*/
def canAdd(expr: Expr[F]): Boolean = {
def checkFor(tree: Tree[F]): Boolean =
tree.head.companion match {
case indented: AndIndented =>
indented.validChildren.map {
case ll: LazyLexem => ll.c
case vc => vc
}.contains(expr.companion)
case _: Prefix =>
tree.tail.value.headOption.exists(checkFor)
case _ => false
}
checkFor(block)
}
/**
* Add child to the block
*/
def add(child: Tree[F]): Block[F] =
copy(content = content :+ child)
/**
* Check if the block has no children
*/
def isEmpty: Boolean = content.isEmpty
/**
* Create a tree corresponding to the block
*/
def close: Tree[F] = {
/**
* Set children of the rightmost expression in tree
*/
def setLast(tree: Tree[F], children: Chain[Tree[F]]): Tree[F] =
tree.copy(tail = tree.tail.map {
case init :== last =>
init :+ setLast(last, children)
case _ =>
children
})
setLast(block, content)
}
}
// TODO(LNG-150): This way of counting indent does not play way with mixed tabs and spaces
private def indentValue[F[_]](indent: F[String])(using Comonad[F]): Int =
indent.extract.length
private def isBlock[F[_]](line: Tree[F]): Boolean =
line.tail.value.headOption
.map(_.head.isBlock)
.getOrElse(line.head.isBlock)
@scala.annotation.tailrec
private def lastExpr[F[_]](tree: Tree[F]): Expr[F] =
tree.tailForced.lastOption match {
case Some(t) => lastExpr(t)
case _ => tree.head
}
def wrongChildError[F[_]](indent: F[String], expr: Expr[F]): ParserError[F] = {
val msg = expr match {
case ReturnExpr(_) =>
"Return expression must be on the top indentation level and at the end of function body"
// could there be other expressions?
case _ => "This expression is on the wrong indentation level"
}
BlockIndentError(indent, msg)
}
def emptyBlockError[F[_]](indent: F[String]): ParserError[F] =
BlockIndentError(indent, "Block expression has no body")
def unexpectedIndentError[F[_]](indent: F[String]): ParserError[F] =
BlockIndentError(indent, "Unexpected indentation")
}