[feature](datatype) add BE config to allow zero date (#34961)

Co-authored-by: Gabriel <[email protected]>

be/src/common/config.cpp

@@ -1061,6 +1061,7 @@ DEFINE_mBool(enable_delete_when_cumu_compaction, "false");
 // max_write_buffer_number for rocksdb
 DEFINE_Int32(rocksdb_max_write_buffer_number, "5");
 
+DEFINE_mBool(allow_zero_date, "false");
 DEFINE_Bool(allow_invalid_decimalv2_literal, "false");
 DEFINE_mString(kerberos_ccache_path, "");
 DEFINE_mString(kerberos_krb5_conf_path, "/etc/krb5.conf");

be/src/common/config.h

@@ -1102,6 +1102,8 @@ DECLARE_mBool(enable_delete_when_cumu_compaction);
 // max_write_buffer_number for rocksdb
 DECLARE_Int32(rocksdb_max_write_buffer_number);
 
+// Convert date 0000-00-00 to 0000-01-01. It's recommended to set to false.
+DECLARE_mBool(allow_zero_date);
 // Allow invalid decimalv2 literal for compatible with old version. Recommend set it false strongly.
 DECLARE_mBool(allow_invalid_decimalv2_literal);
 // Allow to specify kerberos credentials cache path.

be/src/vec/io/io_helper.h

@@ -343,7 +343,7 @@ template <typename T>
 bool read_date_v2_text_impl(T& x, ReadBuffer& buf) {
     static_assert(std::is_same_v<UInt32, T>);
     auto dv = binary_cast<UInt32, DateV2Value<DateV2ValueType>>(x);
-    auto ans = dv.from_date_str(buf.position(), buf.count());
+    auto ans = dv.from_date_str(buf.position(), buf.count(), config::allow_zero_date);
 
     // only to match the is_all_read() check to prevent return null
     buf.position() = buf.end();
@@ -355,7 +355,8 @@ template <typename T>
 bool read_date_v2_text_impl(T& x, ReadBuffer& buf, const cctz::time_zone& local_time_zone) {
     static_assert(std::is_same_v<UInt32, T>);
     auto dv = binary_cast<UInt32, DateV2Value<DateV2ValueType>>(x);
-    auto ans = dv.from_date_str(buf.position(), buf.count(), local_time_zone);
+    auto ans =
+            dv.from_date_str(buf.position(), buf.count(), local_time_zone, config::allow_zero_date);
 
     // only to match the is_all_read() check to prevent return null
     buf.position() = buf.end();
@@ -367,7 +368,7 @@ template <typename T>
 bool read_datetime_v2_text_impl(T& x, ReadBuffer& buf, UInt32 scale = -1) {
     static_assert(std::is_same_v<UInt64, T>);
     auto dv = binary_cast<UInt64, DateV2Value<DateTimeV2ValueType>>(x);
-    auto ans = dv.from_date_str(buf.position(), buf.count(), scale);
+    auto ans = dv.from_date_str(buf.position(), buf.count(), scale, config::allow_zero_date);
 
     // only to match the is_all_read() check to prevent return null
     buf.position() = buf.end();
@@ -380,7 +381,8 @@ bool read_datetime_v2_text_impl(T& x, ReadBuffer& buf, const cctz::time_zone& lo
                                 UInt32 scale = -1) {
     static_assert(std::is_same_v<UInt64, T>);
     auto dv = binary_cast<UInt64, DateV2Value<DateTimeV2ValueType>>(x);
-    auto ans = dv.from_date_str(buf.position(), buf.count(), local_time_zone, scale);
+    auto ans = dv.from_date_str(buf.position(), buf.count(), local_time_zone, scale,
+                                config::allow_zero_date);
 
     // only to match the is_all_read() check to prevent return null
     buf.position() = buf.end();

be/src/vec/runtime/vdatetime_value.cpp

@@ -1974,18 +1974,20 @@ void DateV2Value<T>::format_datetime(uint32_t* date_val, bool* carry_bits) const
 // YYYY-MM-DD HH-MM-DD.FFFFFF AM in default format
 // 0    1  2  3  4  5  6      7
 template <typename T>
-bool DateV2Value<T>::from_date_str(const char* date_str, int len, int scale /* = -1*/) {
-    return from_date_str_base(date_str, len, scale, nullptr);
+bool DateV2Value<T>::from_date_str(const char* date_str, int len, int scale /* = -1*/,
+                                   bool convert_zero) {
+    return from_date_str_base(date_str, len, scale, nullptr, convert_zero);
 }
 // when we parse
 template <typename T>
 bool DateV2Value<T>::from_date_str(const char* date_str, int len,
-                                   const cctz::time_zone& local_time_zone, int scale /* = -1*/) {
-    return from_date_str_base(date_str, len, scale, &local_time_zone);
+                                   const cctz::time_zone& local_time_zone, int scale /* = -1*/,
+                                   bool convert_zero) {
+    return from_date_str_base(date_str, len, scale, &local_time_zone, convert_zero);
 }
 template <typename T>
 bool DateV2Value<T>::from_date_str_base(const char* date_str, int len, int scale,
-                                        const cctz::time_zone* local_time_zone) {
+                                        const cctz::time_zone* local_time_zone, bool convert_zero) {
     const char* ptr = date_str;
     const char* end = date_str + len;
     // ONLY 2, 6 can follow by a space
@@ -2187,7 +2189,12 @@ bool DateV2Value<T>::from_date_str_base(const char* date_str, int len, int scale
         return false;
     }
     if (is_invalid(date_val[0], date_val[1], date_val[2], 0, 0, 0, 0)) {
-        return false;
+        if (date_val[0] == 0 && date_val[1] == 0 && date_val[2] == 0 && convert_zero) {
+            date_val[1] = 1;
+            date_val[2] = 1;
+        } else {
+            return false;
+        }
     }
 
     // In check_range_and_set_time, for Date type the time part will be truncated. So if the timezone offset should make

be/src/vec/runtime/vdatetime_value.h

@@ -847,9 +847,9 @@ class DateV2Value {
     // 'YYMMDD', 'YYYYMMDD', 'YYMMDDHHMMSS', 'YYYYMMDDHHMMSS'
     // 'YY-MM-DD', 'YYYY-MM-DD', 'YY-MM-DD HH.MM.SS'
     // 'YYYYMMDDTHHMMSS'
-    bool from_date_str(const char* str, int len, int scale = -1);
+    bool from_date_str(const char* str, int len, int scale = -1, bool convert_zero = false);
     bool from_date_str(const char* str, int len, const cctz::time_zone& local_time_zone,
-                       int scale = -1);
+                       int scale = -1, bool convert_zero = false);
 
     // Convert this value to string
     // this will check type to decide which format to convert
@@ -1245,7 +1245,7 @@ class DateV2Value {
                              bool disable_lut = false);
 
     bool from_date_str_base(const char* date_str, int len, int scale,
-                            const cctz::time_zone* local_time_zone);
+                            const cctz::time_zone* local_time_zone, bool convert_zero);
 
     // Used to construct from int value
     int64_t standardize_timevalue(int64_t value);

be/test/vec/exprs/vexpr_test.cpp

@@ -519,6 +519,30 @@ TEST(TEST_VEXPR, LITERALTEST) {
         EXPECT_EQ(v, dt);
         EXPECT_EQ("2021-04-07", literal.value());
     }
+    {
+        DateV2Value<DateV2ValueType> data_time_value;
+        const char* date = "00000000";
+        EXPECT_EQ(data_time_value.from_date_str(date, strlen(date), -1, true), true);
+
+        DateV2Value<DateV2ValueType> data_time_value1;
+        const char* date1 = "00000101";
+        EXPECT_EQ(data_time_value1.from_date_str(date1, strlen(date1), -1, true), true);
+        EXPECT_EQ(data_time_value.to_int64(), data_time_value1.to_int64());
+
+        EXPECT_EQ(data_time_value.from_date_str(date, strlen(date)), false);
+    }
+    {
+        DateV2Value<DateTimeV2ValueType> data_time_value;
+        const char* date = "00000000111111";
+        EXPECT_EQ(data_time_value.from_date_str(date, strlen(date), -1, true), true);
+
+        DateV2Value<DateTimeV2ValueType> data_time_value1;
+        const char* date1 = "00000101111111";
+        EXPECT_EQ(data_time_value1.from_date_str(date1, strlen(date1), -1, true), true);
+        EXPECT_EQ(data_time_value.to_int64(), data_time_value1.to_int64());
+
+        EXPECT_EQ(data_time_value.from_date_str(date, strlen(date)), false);
+    }
     // jsonb
     {
         std::string j = R"([null,true,false,100,6.18,"abc"])";