From ea92838cf318a5a6e8f9334ccdef063470cc4d96 Mon Sep 17 00:00:00 2001
From: James Westman <james@jwestman.net>
Date: Thu, 28 Sep 2023 14:41:12 -0500
Subject: [PATCH] Parse escape sequences instead of using replace

That way we can warn about invalid sequences. Also, the previous code had at least one subtle bug (`\\\\'`).
---
 blueprintcompiler/decompiler.py             | 11 +----
 blueprintcompiler/parse_tree.py             | 22 +++++----
 blueprintcompiler/utils.py                  | 52 +++++++++++++++++++++
 tests/sample_errors/bad_escape_sequence.blp |  5 ++
 tests/sample_errors/bad_escape_sequence.err |  1 +
 tests/samples/strings.blp                   |  2 +-
 tests/samples/strings.ui                    |  4 +-
 7 files changed, 76 insertions(+), 21 deletions(-)
 create mode 100644 tests/sample_errors/bad_escape_sequence.blp
 create mode 100644 tests/sample_errors/bad_escape_sequence.err

diff --git a/blueprintcompiler/decompiler.py b/blueprintcompiler/decompiler.py
index 98e00af..550f60b 100644
--- a/blueprintcompiler/decompiler.py
+++ b/blueprintcompiler/decompiler.py
@@ -23,7 +23,7 @@ from dataclasses import dataclass
 from enum import Enum
 
 from .gir import *
-from .utils import Colors
+from .utils import Colors, escape_quote
 from .xml_reader import Element, parse, parse_string
 
 __all__ = ["decompile"]
@@ -253,15 +253,6 @@ def decompiler(tag, cdata=False):
     return decorator
 
 
-def escape_quote(string: str) -> str:
-    return (
-        string.replace("\\", "\\\\")
-        .replace("'", "\\'")
-        .replace('"', '\\"')
-        .replace("\n", "\\n")
-    )
-
-
 @decompiler("interface")
 def decompile_interface(ctx, gir):
     return gir
diff --git a/blueprintcompiler/parse_tree.py b/blueprintcompiler/parse_tree.py
index 8f3ef31..fff6e4a 100644
--- a/blueprintcompiler/parse_tree.py
+++ b/blueprintcompiler/parse_tree.py
@@ -22,6 +22,7 @@
 import typing as T
 from enum import Enum
 
+from . import utils
 from .ast_utils import AstNode
 from .errors import (
     CompileError,
@@ -573,14 +574,19 @@ class UseQuoted(ParseNode):
         if token.type != TokenType.QUOTED:
             return False
 
-        string = (
-            str(token)[1:-1]
-            .replace("\\n", "\n")
-            .replace('\\"', '"')
-            .replace("\\\\", "\\")
-            .replace("\\'", "'")
-        )
-        ctx.set_group_val(self.key, string, token)
+        unescaped = None
+
+        try:
+            unescaped = utils.unescape_quote(str(token))
+        except utils.UnescapeError as e:
+            start = ctx.tokens[ctx.index - 1].start
+            range = Range(start + e.start, start + e.end, ctx.text)
+            ctx.errors.append(
+                CompileError(f"Invalid escape sequence '{range.text}'", range)
+            )
+
+        ctx.set_group_val(self.key, unescaped, token)
+
         return True
 
 
diff --git a/blueprintcompiler/utils.py b/blueprintcompiler/utils.py
index 4c4b44a..6d03a9e 100644
--- a/blueprintcompiler/utils.py
+++ b/blueprintcompiler/utils.py
@@ -18,6 +18,7 @@
 # SPDX-License-Identifier: LGPL-3.0-or-later
 
 import typing as T
+from dataclasses import dataclass
 
 
 class Colors:
@@ -98,3 +99,54 @@ def idxs_to_range(start: int, end: int, text: str):
             "character": end_c,
         },
     }
+
+
+@dataclass
+class UnescapeError(Exception):
+    start: int
+    end: int
+
+
+def escape_quote(string: str) -> str:
+    return (
+        string.replace("\\", "\\\\")
+        .replace("'", "\\'")
+        .replace('"', '\\"')
+        .replace("\n", "\\n")
+        .replace("\t", "\\t")
+    )
+
+
+def unescape_quote(string: str) -> str:
+    string = string[1:-1]
+
+    REPLACEMENTS = {
+        "\\": "\\",
+        "n": "\n",
+        "t": "\t",
+        '"': '"',
+        "'": "'",
+    }
+
+    result = ""
+    i = 0
+    while i < len(string):
+        c = string[i]
+        if c == "\\":
+            i += 1
+
+            if i >= len(string):
+                from .errors import CompilerBugError
+
+                raise CompilerBugError()
+
+            if r := REPLACEMENTS.get(string[i]):
+                result += r
+            else:
+                raise UnescapeError(i, i + 2)
+        else:
+            result += c
+
+        i += 1
+
+    return result
diff --git a/tests/sample_errors/bad_escape_sequence.blp b/tests/sample_errors/bad_escape_sequence.blp
new file mode 100644
index 0000000..4b109c6
--- /dev/null
+++ b/tests/sample_errors/bad_escape_sequence.blp
@@ -0,0 +1,5 @@
+using Gtk 4.0;
+
+Label {
+  label: '***** \f *****';
+}
diff --git a/tests/sample_errors/bad_escape_sequence.err b/tests/sample_errors/bad_escape_sequence.err
new file mode 100644
index 0000000..e4ec183
--- /dev/null
+++ b/tests/sample_errors/bad_escape_sequence.err
@@ -0,0 +1 @@
+4,17,2,Invalid escape sequence '\f'
\ No newline at end of file
diff --git a/tests/samples/strings.blp b/tests/samples/strings.blp
index ef237ae..08bb418 100644
--- a/tests/samples/strings.blp
+++ b/tests/samples/strings.blp
@@ -1,5 +1,5 @@
 using Gtk 4.0;
 
 Label {
-  label: "Test 1 2 3\n & 4 \"5\' 6";
+  label: "\\\\'Test 1 2 3\n & 4 \"5\' 6 \t";
 }
diff --git a/tests/samples/strings.ui b/tests/samples/strings.ui
index 1dea963..e7fef2e 100644
--- a/tests/samples/strings.ui
+++ b/tests/samples/strings.ui
@@ -7,7 +7,7 @@ corresponding .blp file and regenerate this file with blueprint-compiler.
 <interface>
   <requires lib="gtk" version="4.0"/>
   <object class="GtkLabel">
-    <property name="label">Test 1 2 3
- &amp; 4 "5' 6</property>
+    <property name="label">\\'Test 1 2 3
+ &amp; 4 "5' 6 	</property>
   </object>
 </interface>