fix CONNECT and PRIOR keywords in Oracle syntax

Author: richarah

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.21)
 cmake_policy(SET CMP0048 NEW)
-project(libsqlglot VERSION 0.2.10 LANGUAGES CXX)
+project(libsqlglot VERSION 0.3.1 LANGUAGES CXX)
 
 # Generate version header
 configure_file(

README.md

@@ -1,16 +1,20 @@
 # libsqlglot
 
-#### What is this?
+#### What this is
 
 sqlglot, in C++. 45 dialects, 126× faster on benchmark SQL, 252× on the kind your ORM generates when nobody's looking. Performance gap scales with query complexity, see [Benchmarks](#benchmarks).
 
 Supports stored procedures (PL/pgSQL, T-SQL, MySQL, PL/SQL): where sqlglot falls back to passthrough, libsqlglot parses them into the AST.
 
 Inspired by the original [sqlglot](https://github.com/tobymao/sqlglot), which did the decade-long work of mapping 31+ SQL dialects into an elegant, universal AST. libsqlglot does the comparatively trivial work of compiling it. The algorithm was already O(n), the runtime was O(python).
 
-#### What is this for?
+#### What it's for
 
-DB proxies, linters, migration tools, query rewriters, and a proper replacement for every regex that's pretending to parse SQL.
+Proxies, sidecars, migration tools, linters, anything where SQL hits a hot path. Also a proper replacement for every regex that's pretending to parse queries.
+
+#### Why it exists
+
+Because the highest praise one can pay a Python 
 
 ## Contents
 

dialects/patches/oracle.json

@@ -4,6 +4,8 @@
   "description": "Oracle Database - Enterprise RDBMS with PL/SQL",
   "keywords": {
     "add": [
+      {"name": "CONNECT", "token_type": "CONNECT"},
+      {"name": "PRIOR", "token_type": "PRIOR"},
       {"name": "FLASHBACK", "token_type": "FLASHBACK"},
       {"name": "VERSIONS", "token_type": "VERSIONS"},
       {"name": "SCN", "token_type": "SCN"},

include/libsqlglot/tokenizer.h

@@ -86,6 +86,9 @@ class Tokenizer {
             return '\0';  // Out of bounds
         }
         size_t p = pos_ + offset;
+        if (p >= source_.size()) {
+            return '\0';  // Out of bounds
+        }
         return source_[p];
     }