Merge pull request #55035 from rschu1ze/is_deterministic_resubmit

Added field "is_deterministic" to system.functions
This commit is contained in:
Robert Schulze 2023-09-27 10:13:18 +02:00 committed by GitHub
commit 3a1663a460
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 85 additions and 36 deletions

View File

@ -7,28 +7,34 @@ Contains information about normal and aggregate functions.
Columns:
- `name`(`String`) The name of the function.
- `is_aggregate`(`UInt8`) — Whether the function is aggregate.
- `name` ([String](../../sql-reference/data-types/string.md)) The name of the function.
- `is_aggregate` ([UInt8](../../sql-reference/data-types/int-uint.md)) — Whether the function is an aggregate function.
- `is_deterministic` ([Nullable](../../sql-reference/data-types/nullable.md)([UInt8](../../sql-reference/data-types/int-uint.md))) - Whether the function is deterministic.
- `case_insensitive`, ([UInt8](../../sql-reference/data-types/int-uint.md)) - Whether the function name can be used case-insensitively.
- `alias_to`, ([String](../../sql-reference/data-types/string.md)) - The original function name, if the function name is an alias.
- `create_query`, ([String](../../sql-reference/data-types/enum.md)) - Unused.
- `origin`, ([Enum8](../../sql-reference/data-types/string.md)) - Unused.
- `description`, ([String](../../sql-reference/data-types/string.md)) - A high-level description what the function does.
- `syntax`, ([String](../../sql-reference/data-types/string.md)) - Signature of the function.
- `arguments`, ([String](../../sql-reference/data-types/string.md)) - What arguments does the function take.
- `returned_value`, ([String](../../sql-reference/data-types/string.md)) - What does the function return.
- `examples`, ([String](../../sql-reference/data-types/string.md)) - Example usage of the function.
- `categories`, ([String](../../sql-reference/data-types/string.md)) - The category of the function.
**Example**
```sql
SELECT * FROM system.functions LIMIT 10;
SELECT name, is_aggregate, is_deterministic, case_insensitive, alias_to FROM system.functions LIMIT 5;
```
```text
┌─name──────────────────┬─is_aggregate─┬─case_insensitive─┬─alias_to─┬─create_query─┬─origin─┐
│ logTrace │ 0 │ 0 │ │ │ System │
│ aes_decrypt_mysql │ 0 │ 0 │ │ │ System │
│ aes_encrypt_mysql │ 0 │ 0 │ │ │ System │
│ decrypt │ 0 │ 0 │ │ │ System │
│ encrypt │ 0 │ 0 │ │ │ System │
│ toBool │ 0 │ 0 │ │ │ System │
│ windowID │ 0 │ 0 │ │ │ System │
│ hopStart │ 0 │ 0 │ │ │ System │
│ hop │ 0 │ 0 │ │ │ System │
│ snowflakeToDateTime64 │ 0 │ 0 │ │ │ System │
└───────────────────────┴──────────────┴──────────────────┴──────────┴──────────────┴────────┘
┌─name─────────────────────┬─is_aggregate─┬─is_deterministic─┬─case_insensitive─┬─alias_to─┐
│ BLAKE3 │ 0 │ 1 │ 0 │ │
│ sipHash128Reference │ 0 │ 1 │ 0 │ │
│ mapExtractKeyLike │ 0 │ 1 │ 0 │ │
│ sipHash128ReferenceKeyed │ 0 │ 1 │ 0 │ │
│ mapPartialSort │ 0 │ 1 │ 0 │ │
└──────────────────────────┴──────────────┴──────────────────┴──────────────────┴──────────┘
10 rows in set. Elapsed: 0.002 sec.
5 rows in set. Elapsed: 0.002 sec.
```

View File

@ -1,5 +1,6 @@
#include <AggregateFunctions/AggregateFunctionFactory.h>
#include <DataTypes/DataTypeMap.h>
#include <DataTypes/DataTypeNullable.h>
#include <DataTypes/DataTypeString.h>
#include <DataTypes/DataTypesNumber.h>
#include <DataTypes/DataTypeEnum.h>
@ -15,6 +16,15 @@
namespace DB
{
namespace ErrorCodes
{
extern const int DICTIONARIES_WAS_NOT_LOADED;
extern const int FUNCTION_NOT_ALLOWED;
extern const int LOGICAL_ERROR;
extern const int NOT_IMPLEMENTED;
extern const int SUPPORT_IS_DISABLED;
};
enum class FunctionOrigin : Int8
{
SYSTEM = 0,
@ -29,6 +39,7 @@ namespace
MutableColumns & res_columns,
const String & name,
UInt64 is_aggregate,
std::optional<UInt64> is_deterministic,
const String & create_query,
FunctionOrigin function_origin,
const Factory & factory)
@ -36,53 +47,58 @@ namespace
res_columns[0]->insert(name);
res_columns[1]->insert(is_aggregate);
if (!is_deterministic.has_value())
res_columns[2]->insertDefault();
else
res_columns[2]->insert(*is_deterministic);
if constexpr (std::is_same_v<Factory, UserDefinedSQLFunctionFactory> || std::is_same_v<Factory, UserDefinedExecutableFunctionFactory>)
{
res_columns[2]->insert(false);
res_columns[3]->insertDefault();
res_columns[3]->insert(false);
res_columns[4]->insertDefault();
}
else
{
res_columns[2]->insert(factory.isCaseInsensitive(name));
res_columns[3]->insert(factory.isCaseInsensitive(name));
if (factory.isAlias(name))
res_columns[3]->insert(factory.aliasTo(name));
res_columns[4]->insert(factory.aliasTo(name));
else
res_columns[3]->insertDefault();
res_columns[4]->insertDefault();
}
res_columns[4]->insert(create_query);
res_columns[5]->insert(static_cast<Int8>(function_origin));
res_columns[5]->insert(create_query);
res_columns[6]->insert(static_cast<Int8>(function_origin));
if constexpr (std::is_same_v<Factory, FunctionFactory>)
{
if (factory.isAlias(name))
{
res_columns[6]->insertDefault();
res_columns[7]->insertDefault();
res_columns[8]->insertDefault();
res_columns[9]->insertDefault();
res_columns[10]->insertDefault();
res_columns[11]->insertDefault();
res_columns[12]->insertDefault();
}
else
{
auto documentation = factory.getDocumentation(name);
res_columns[6]->insert(documentation.description);
res_columns[7]->insert(documentation.syntax);
res_columns[8]->insert(documentation.argumentsAsString());
res_columns[9]->insert(documentation.returned_value);
res_columns[10]->insert(documentation.examplesAsString());
res_columns[11]->insert(documentation.categoriesAsString());
res_columns[7]->insert(documentation.description);
res_columns[8]->insert(documentation.syntax);
res_columns[9]->insert(documentation.argumentsAsString());
res_columns[10]->insert(documentation.returned_value);
res_columns[11]->insert(documentation.examplesAsString());
res_columns[12]->insert(documentation.categoriesAsString());
}
}
else
{
res_columns[6]->insertDefault();
res_columns[7]->insertDefault();
res_columns[8]->insertDefault();
res_columns[9]->insertDefault();
res_columns[10]->insertDefault();
res_columns[11]->insertDefault();
res_columns[12]->insertDefault();
}
}
}
@ -102,6 +118,7 @@ NamesAndTypesList StorageSystemFunctions::getNamesAndTypes()
return {
{"name", std::make_shared<DataTypeString>()},
{"is_aggregate", std::make_shared<DataTypeUInt8>()},
{"is_deterministic", std::make_shared<DataTypeNullable>(std::make_shared<DataTypeUInt8>())},
{"case_insensitive", std::make_shared<DataTypeUInt8>()},
{"alias_to", std::make_shared<DataTypeString>()},
{"create_query", std::make_shared<DataTypeString>()},
@ -121,14 +138,34 @@ void StorageSystemFunctions::fillData(MutableColumns & res_columns, ContextPtr c
const auto & function_names = functions_factory.getAllRegisteredNames();
for (const auto & function_name : function_names)
{
fillRow(res_columns, function_name, UInt64(0), "", FunctionOrigin::SYSTEM, functions_factory);
std::optional<UInt64> is_deterministic;
try
{
is_deterministic = functions_factory.tryGet(function_name, context)->isDeterministic();
}
catch (const Exception & e)
{
/// Some functions throw because they need special configuration or setup before use.
if (e.code() == ErrorCodes::DICTIONARIES_WAS_NOT_LOADED
|| e.code() == ErrorCodes::FUNCTION_NOT_ALLOWED
|| e.code() == ErrorCodes::LOGICAL_ERROR
|| e.code() == ErrorCodes::NOT_IMPLEMENTED
|| e.code() == ErrorCodes::SUPPORT_IS_DISABLED)
{
/// Ignore exception, show is_deterministic = NULL.
}
else
throw;
}
fillRow(res_columns, function_name, 0, is_deterministic, "", FunctionOrigin::SYSTEM, functions_factory);
}
const auto & aggregate_functions_factory = AggregateFunctionFactory::instance();
const auto & aggregate_function_names = aggregate_functions_factory.getAllRegisteredNames();
for (const auto & function_name : aggregate_function_names)
{
fillRow(res_columns, function_name, UInt64(1), "", FunctionOrigin::SYSTEM, aggregate_functions_factory);
fillRow(res_columns, function_name, 1, {1}, "", FunctionOrigin::SYSTEM, aggregate_functions_factory);
}
const auto & user_defined_sql_functions_factory = UserDefinedSQLFunctionFactory::instance();
@ -136,14 +173,14 @@ void StorageSystemFunctions::fillData(MutableColumns & res_columns, ContextPtr c
for (const auto & function_name : user_defined_sql_functions_names)
{
auto create_query = queryToString(user_defined_sql_functions_factory.get(function_name));
fillRow(res_columns, function_name, UInt64(0), create_query, FunctionOrigin::SQL_USER_DEFINED, user_defined_sql_functions_factory);
fillRow(res_columns, function_name, 0, {0}, create_query, FunctionOrigin::SQL_USER_DEFINED, user_defined_sql_functions_factory);
}
const auto & user_defined_executable_functions_factory = UserDefinedExecutableFunctionFactory::instance();
const auto & user_defined_executable_functions_names = user_defined_executable_functions_factory.getRegisteredNames(context);
for (const auto & function_name : user_defined_executable_functions_names)
{
fillRow(res_columns, function_name, UInt64(0), "", FunctionOrigin::EXECUTABLE_USER_DEFINED, user_defined_executable_functions_factory);
fillRow(res_columns, function_name, 0, {0}, "", FunctionOrigin::EXECUTABLE_USER_DEFINED, user_defined_executable_functions_factory);
}
}

View File

@ -283,6 +283,7 @@ CREATE TABLE system.functions
(
`name` String,
`is_aggregate` UInt8,
`is_deterministic` Nullable(UInt8),
`case_insensitive` UInt8,
`alias_to` String,
`create_query` String,

View File

@ -53,4 +53,9 @@ expect eof
EOF
}
run "$CLICKHOUSE_LOCAL"
run "$CLICKHOUSE_LOCAL --disable_suggestion"
# Suggestions are off because the suggestion feature initializes itself by reading all available function
# names from "system.functions". Getting the value for field "is_obsolete" occasionally throws (e.g. for
# certain dictionary functions when dictionaries are not set up yet). Exceptions are properly handled, but
# they exist for a short time. This, in combination with CLICKHOUSE_TERMINATE_ON_ANY_EXCEPTION, terminates
# clickhouse-local and clickhouse-client when run in interactive mode *with* suggestions.