improve error message

examples/evalexpr.py

@@ -7,6 +7,7 @@ import sys
 
 sys.path.append(f"{dirname(__file__)}/../src")
 from pyalibert import regex, Parser, just, Declare
+from pyalibert.result import ParseError
 
 # ast modeling
 
@@ -79,9 +80,12 @@ def repl():
     print("> ", end="")
     sys.stdout.flush()
     for line in sys.stdin:
-        result = parser.parse(line.strip())
-        print(result)
-        print(f"= {result.eval()}")
+        try:
+            result = parser.parse(line.strip())
+            print(result)
+            print(f"= {result.eval()}")
+        except ParseError as err:
+            print(err)
         print("> ", end="")
         sys.stdout.flush()
 

src/pyalibert/result.py

@@ -1,7 +1,7 @@
 from dataclasses import dataclass
-from typing import Union, Generic, TypeVar
+from typing import Generator, Union, Generic, TypeVar
 
-from .utils import TypeInfo
+from .utils import best_by_key, last, split_lines_into_bounded
 
 
 P = TypeVar("P")
@@ -16,13 +16,38 @@ class Success(Generic[P]):
 class Failure:
     at_index: int
     expected: set[str]
-    depth: int | None = None
+    children: list["Failure"]
+
+    def __init__(self, at_index: int, expected: set[str], *children: "Failure") -> None:
+        self.at_index = at_index
+        self.expected = expected
+        for child in children:
+            for e in child.expected:
+                self.expected.add(e)
+        self.children = [*children]
+
+    def depth(self):
+        depths = [child.depth() + 1 for child in self.children]
+        return max(0, 0, *depths)
+
+    def deepest_descendency(self):
+        def generator(failure: Failure) -> Generator[Failure, None, None]:
+            deepest = best_by_key(failure.children, lambda f: f.depth())
+            if deepest is None:
+                return
+            yield deepest
+            yield from generator(deepest)
+        return [*generator(self)]
+
+    def parent(self):
+        return Failure(self.at_index, self.expected, self)
 
 
 P = TypeVar("P")
 Result = Union[Failure, Success[P]]
 
 
+@dataclass
 class ParseError(BaseException):
     failure: Failure
     stream: str
@@ -30,11 +55,36 @@ class ParseError(BaseException):
     def __init__(self, failure: Failure, stream: str) -> None:
         self.failure = failure
         self.stream = stream
-
-        fail_section = stream[failure.at_index:80]
-        message = f"""
-Parsing failed at position {self.failure.at_index} of stream :
-${fail_section}
-Expected one of: {failure.expected}
-""".strip()
+        message = self.format_error_secton()
         super().__init__(message)
+
+    def format_error_secton(self):
+        actual_failure = last(self.failure.deepest_descendency()) or self.failure
+        failure_pos = actual_failure.at_index
+        lines = [*split_lines_into_bounded(self.stream)]
+        failure_line_index = next(i for (i, ((s, e), l)) in enumerate(lines) if s <= failure_pos and e >= failure_pos)
+        result = "\n"
+        if failure_line_index > 0:
+            index = failure_line_index - 1
+            result += format_error_section_line(index, lines[index][1])[1] + "\n"
+        (prefix_len, line) = format_error_section_line(failure_line_index, lines[failure_line_index][1])
+        result += line + "\n"
+        result += format_error_section_indicator(failure_pos, lines[failure_line_index][0][0], prefix_len, actual_failure.expected) + "\n"
+        if failure_line_index < (len(lines) - 1):
+            index = failure_line_index + 1
+            result += format_error_section_line(index, lines[index][1])[1]
+        return result
+
+
+def format_error_section_line(index: int, line: str):
+    line_num_part = str(index).rjust(4)
+    prefix = f"{line_num_part} |"
+    result = f"{prefix}{line}"
+    return (len(prefix), result)
+
+
+def format_error_section_indicator(failure_pos: int, failure_line_start_pos: int, formatted_line_prefix_len: int, expected: set[str]):
+    relative_pos = failure_pos - failure_line_start_pos
+    padding = formatted_line_prefix_len + relative_pos
+    expects = ", ".join(e for e in expected)
+    return f"{' ' * padding}^\\_ expected : {expects}"

src/pyalibert/transformers/and_.py

@@ -15,10 +15,10 @@ class AndTransform(Transformer[tuple[L, R]]):
     def parse(self, stream: str, at_index: int) -> Result[tuple[L, R]]:
         result_left = self.left.parse(stream, at_index)
         if isinstance(result_left, Failure):
-            return result_left
+            return result_left.parent()
         result_right = self.right.parse(stream, result_left.next_index)
         if isinstance(result_right, Failure):
-            return result_right
+            return result_right.parent()
         return Success((result_left.value, result_right.value), result_right.next_index)
 
 
@@ -42,4 +42,4 @@ def test_and_shifted(ctx):
 def test_and_none(ctx):
     parser = AndTransform(JustTransform("ar"), JustTransform("bre"))
     input = "......"
-    assert parser.parse(input, 2) == Failure(2, set(["ar"]))
+    assert parser.parse(input, 2) == Failure(2, set(["ar"]), Failure(2, set(["ar"])))

src/pyalibert/transformers/or_.py

@@ -19,7 +19,7 @@ class OrTransform(Transformer[Union[L, R]]):
         result_right = self.right.parse(stream, at_index)
         if isinstance(result_right, Success):
             return Success(result_right.value, result_right.next_index)
-        return Failure(at_index, result_left.expected.union(result_right.expected))
+        return Failure(at_index, set(), result_left, result_right)
 
 
 from okipy.lib import test
@@ -48,4 +48,4 @@ def test_or_shifted(ctx):
 def test_or_fail(ctx):
     parser = OrTransform(JustTransform("ar"), JustTransform("bre"))
     input = "......"
-    assert parser.parse(input, 2) == Failure(2, set(["ar", "bre"]))
+    assert parser.parse(input, 2) == Failure(2, set(["ar", "bre"]), Failure(2, set(["ar"])), Failure(2, set(["bre"])))

src/pyalibert/utils.py

@@ -1,5 +1,59 @@
-from typing import Generic, TypeVar
+from typing import Callable, Collection, Generator, Generic, Iterable, TypeVar, Union
+
+from okipy.lib import test
+
 
 T = TypeVar("T")
 class TypeInfo(Generic[T]):
     pass
+
+
+T = TypeVar("T")
+K = TypeVar("K", int, float)
+def best_by_key(collection: Iterable[T], extract: Callable[[T], K], max_: bool = True):
+    best: Union[tuple[T,K], None] = None
+    for item in collection:
+        key = extract(item)
+        if best is None:
+            best = (item, key)
+            continue
+        should_replace = (max_ and key > best[1]) or (not max_ and key < best[1])
+        if should_replace:
+            best = (item, key)
+    return best[0] if best is not None else None
+
+
+def split_lines_into_bounded(text: str) -> Generator[tuple[tuple[int, int], str], None, None]:
+    rest = text
+    acc = 0
+    while True:
+        parts = rest.split("\n", 1)
+        start = acc
+        acc += len(parts[0])
+        end = acc
+        yield ((start, end), parts[0])
+        acc += 1
+        if len(parts) == 1:
+            break
+        rest = parts[1]
+
+
+@test()
+def test_split_lines_bounded(ctx):
+    input = "\nhello\n\nworld\n"
+    splitted = [*split_lines_into_bounded(input)]
+    assert splitted == [
+        ((0, 0), ''),
+        ((1, 6), 'hello'),
+        ((7, 7), ''),
+        ((8, 13), 'world'),
+        ((14, 14), '')
+    ]
+
+
+T = TypeVar("T")
+def last(collection: Collection[T]):
+    result = None
+    for item in collection:
+        result = item
+    return result