tinybird-toolset 2.4.11__tar.gz → 2.4.14__tar.gz

{tinybird_toolset-2.4.11/src/tinybird_toolset.egg-info → tinybird_toolset-2.4.14}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tinybird-toolset
-Version: 2.4.11
+Version: 2.4.14
 Home-page: https://gitlab.com/tinybird/clickhouse-toolset
 Author: Tinybird.co
 Author-email: support@tinybird.co

{tinybird_toolset-2.4.11 → tinybird_toolset-2.4.14}/conf.py

@@ -32,6 +32,24 @@ MSAN_ENABLED = bool(os.environ.get('MSAN_ENABLED', ''))  # MemorySanitizer
 TSAN_ENABLED = bool(os.environ.get('TSAN_ENABLED', ''))  # ThreadSanitizer
 THINLTO_FLAG = '-flto=thin'
 
+# Taken from https://github.com/cpp-best-practices/cmake_template/blob/a3971f5b45/cmake/CompilerWarnings.cmake
+CLANG_WARNINGS = [
+    '-Wall',
+    '-Wextra',
+    '-Wshadow',
+    '-Wnon-virtual-dtor',
+    '-Wold-style-cast',
+    '-Wcast-align',
+    '-Wunused',
+    '-Woverloaded-virtual',
+    '-Wpedantic',
+    '-Wconversion',
+    '-Wnull-dereference',
+    '-Wdouble-promotion',
+    '-Wformat=2',
+    '-Wimplicit-fallthrough',
+]
+
 
 def apply_patch(patch_path):
     if OMIT_PATCHES:
@@ -123,6 +141,11 @@ class ClickHouseBuildExt(build_ext):
         # Temporary for debugging: some functions are missing from libdbms.a, this is in the private repo pmaster already so we should include it from the repo
         apply_patch("patches/dbmsfunctions.patch")
 
+        # Range-check narrow-integer Field inserts so JSON/Dynamic paths widen instead of
+        # silently wrapping (1944 -> -104). Backport of the ColumnVector::tryInsert fix;
+        # remove once the fix lands in the pinned ClickHouse submodule.
+        apply_patch("patches/ColumnVectorTryInsert.patch")
+
         if sys.platform == 'darwin':
             # We disable whole_archive options destined for AppleClang, which is unsupported but keeps being a PITA
             apply_patch("patches/darwin.patch")
@@ -276,10 +299,13 @@ class CustomBuildWithFromCH(build_ext):
     def cflags():
         cflags_list = CustomBuildWithFromCH.ninja_extractor_other('FLAGS')
         include_list = list(map(transform_path, CustomBuildWithFromCH.ninja_extractor_other('INCLUDES')))
-        # Reduce the noise coming from python since those flags are added later as -I and not -isystem
-        other_list = ['-Wno-reserved-identifier', '-Wno-zero-as-null-pointer-constant', '-Wno-old-style-cast',
-                      '-Wno-cast-function-type', '-Wno-unused-function', '-Wno-redundant-decls',
-                      '-Wno-missing-include-dirs']
+        # Replace -I with -isystem: these are external headers and we don't want to see warnings from them
+        include_list = [flag.replace('-I', '-isystem', 1) if flag.startswith('-I') else flag for flag in include_list]
+        # Remove -W flags, set our own warnings
+        cflags_list = [flag for flag in cflags_list if not flag.startswith('-W')]
+        other_list = CLANG_WARNINGS.copy()
+        other_list += ['-Werror']  # Treat warnings as errors
+
         if BUILD_FOR_VALGRIND:
             other_list += ['-g', '-gdwarf-4']
         elif DEBUG_SYMBOLS:
@@ -332,6 +358,10 @@ class CustomBuildWithFromCH(build_ext):
 
         # Remove -Wl,--no-undefined since Python symbols will undefined
         link_flags = filter(lambda flag: flag.find('--no-undefined') == -1, link_flags)
+        if sys.platform == 'darwin':
+            # -undefined dynamic_lookup (added in extra_libs) already permits all undefined
+            # symbols, so an explicit -Wl,-U,<symbol> is redundant and makes ld emit a warning.
+            link_flags = filter(lambda flag: not flag.startswith('-Wl,-U,'), link_flags)
         return " ".join(link_flags)
 
     @staticmethod
@@ -384,7 +414,11 @@ class CustomBuildWithFromCH(build_ext):
         linker = getattr(self.compiler, "linker_so_cxx")
         linker = [COMPILER_CC] + split_quoted(self.ldflags()) + linker[1:]
 
-        compiler += ['-isystem', TOOLSET_PATH]
+        compiler += ['-I', TOOLSET_PATH]
+        # Make include_dirs (Python headers; set by setuptools) to -isystem: prevents warnings from Python headers
+        for d in self.include_dirs or []:
+            compiler += ['-isystem', d]
+        self.include_dirs = []
 
         self.compiler.set_executable("compiler_so", compiler)
         self.compiler.set_executable("compiler_so_cxx", compiler)

{tinybird_toolset-2.4.11 → tinybird_toolset-2.4.14}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, Extension
 
 NAME = 'tinybird-toolset'
-VERSION = '2.4.11'
+VERSION = '2.4.14'
 
 try:
     from conf import *

{tinybird_toolset-2.4.11 → tinybird_toolset-2.4.14/src/tinybird_toolset.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tinybird-toolset
-Version: 2.4.11
+Version: 2.4.14
 Home-page: https://gitlab.com/tinybird/clickhouse-toolset
 Author: Tinybird.co
 Author-email: support@tinybird.co

{tinybird_toolset-2.4.11 → tinybird_toolset-2.4.14}/tests/test_convert_to_row_binary.py

@@ -5134,6 +5134,43 @@ TEST_CASES = [
         b'\x00\x00L-100000000000000000000000000000000000000000000000000000000000000000000000000\x00\x01\x1dInvalid decimal string format\x00\x01\x05value\x01\x01',
         0, 1),
 
+    # Booleans imported into Decimal must produce EXACTLY 1.0 and 0.0, without precision loss
+    (ConversionMode.ONLY_LEGACY,
+        [('t', 'Decimal32(4)', '$.t', False), ('f', 'Decimal32(4)', '$.f', False)],
+        '{"t":true,"f":false}',
+        b'\x00\x10\x27\x00\x00\x00\x00\x00\x00\x00',
+        b'',
+        1, 0),
+
+    (ConversionMode.ONLY_LEGACY,
+        [('t', 'Decimal64(8)', '$.t', False), ('f', 'Decimal64(8)', '$.f', False)],
+        '{"t":true,"f":false}',
+        b'\x00\x00\xe1\xf5\x05\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00',
+        b'',
+        1, 0),
+
+    (ConversionMode.ONLY_LEGACY,
+        [('t', 'Decimal128(30)', '$.t', False), ('f', 'Decimal128(30)', '$.f', False)],
+        '{"t":true,"f":false}',
+        b'\x00\x00\x00\x00\x40\xea\xed\x74\x46\xd0\x9c\x2c\x9f\x0c\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00',
+        b'',
+        1, 0),
+
+    (ConversionMode.ONLY_LEGACY,
+        [('t', 'Decimal256(50)', '$.t', False), ('f', 'Decimal256(50)', '$.f', False)],
+        '{"t":true,"f":false}',
+        b'\x00\x00\x00\x00\x00\x00\x00\x64\xb5\xfd\x34\x05\xc4\xd2\x87\x66\x92\xf9\x15\x3b\x6c\x44\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00',
+        b'',
+        1, 0),
+
+    # Non-legacy mode rejects booleans imported into Decimal columns: the row is quarantined.
+    (ConversionMode.ONLY_NONLEGACY,
+        [('value', 'Decimal128(30)', '$.value', False)],
+        '{"value": true}',
+        b'',
+        b"\x00\x00\x04true\x00\x01XInvalid value true for type 'Decimal(38, 30)' in column 'value' with jsonpath '$.value'.\x00\x01\x05value\x01\x01",
+        0, 1),
+
     (ConversionMode.ALWAYS,
         [('value', 'FixedString(6)', '$.value', False)],
         '{}',
@@ -8309,20 +8346,22 @@ TEST_CASES = [
         1, 0),
 
 
-    # schemaless JSON encoded as string can choose too-narrow integer types and overflow values
+    # schemaless JSON encoded as string widens too-narrow integer types instead of
+    # overflowing: y=22 first pins the path to Int8, then y=220 widens it to Int16
+    # rather than wrapping to -36 (CH-440, ColumnVector::tryInsert range check)
     (ConversionMode.ONLY_LEGACY,
         [('v', 'Map(String, JSON)', '$', False)],
         '{"a":{"x":11,"y":22},"b":{"x":110,"y":220}}',
-        b'\x00\x02\x01a\x0f{"x":11,"y":22}\x01b\x11{"x":110,"y":-36}',
+        b'\x00\x02\x01a\x0f{"x":11,"y":22}\x01b\x11{"x":110,"y":220}',
         b'',
         1, 0),
     (ConversionMode.ONLY_LEGACY,
         [('json', 'Map(String, JSON)', "$", False)],
         '{"a":"{\\"x\\":11,\\"y\\":22}","b":"{\\"x\\":110,\\"y\\":220}"}',
-        b'\x00\x02\x01a\x0f{"x":11,"y":22}\x01b\x11{"x":110,"y":-36}',
+        b'\x00\x02\x01a\x0f{"x":11,"y":22}\x01b\x11{"x":110,"y":220}',
         b'',
         1, 0),
-    # doesn't happen with binary JSON encoding
+    # binary JSON encoding stores per-value type tags, so it already preserved the value
     (ConversionMode.ONLY_NONLEGACY,
         [('v', 'Map(String, JSON)', '$', False)],
         '{"a":{"x":11,"y":22},"b":{"x":110,"y":220}}',