1 | #include "duckdb/function/scalar/string_functions.hpp" |
2 | |
3 | #include "duckdb/common/exception.hpp" |
4 | #include "duckdb/common/vector_operations/vector_operations.hpp" |
5 | #include "duckdb/common/vector_operations/ternary_executor.hpp" |
6 | #include "utf8proc.hpp" |
7 | |
8 | using namespace std; |
9 | |
10 | namespace duckdb { |
11 | |
12 | string_t SubstringFun::substring_ascii_only(Vector &result, const char *input_data, int offset, int length) { |
13 | auto result_string = StringVector::EmptyString(result, length); |
14 | auto result_data = result_string.GetData(); |
15 | memcpy(result_data, input_data + offset, length); |
16 | result_string.Finalize(); |
17 | return result_string; |
18 | } |
19 | |
20 | string_t SubstringFun::substring_scalar_function(Vector &result, string_t input, int offset, int length, |
21 | unique_ptr<char[]> &output, idx_t ¤t_len) { |
22 | // reduce offset by one because SQL starts counting at 1 |
23 | offset--; |
24 | if (offset < 0 || length < 0) { |
25 | throw Exception("SUBSTRING cannot handle negative offsets" ); |
26 | } |
27 | auto input_data = input.GetData(); |
28 | auto input_size = input.GetSize(); |
29 | |
30 | // check if there is any non-ascii |
31 | bool ascii_only = true; |
32 | int ascii_end = std::min(offset + length + 1, (int)input_size); |
33 | for (int i = 0; i < ascii_end; i++) { |
34 | if (input_data[i] & 0x80) { |
35 | ascii_only = false; |
36 | break; |
37 | } |
38 | } |
39 | |
40 | if (length == 0) { |
41 | auto result_string = StringVector::EmptyString(result, 0); |
42 | result_string.Finalize(); |
43 | return result_string; |
44 | } |
45 | |
46 | if (ascii_only) { |
47 | // ascii only |
48 | length = std::min(offset + length, (int)input_size); |
49 | if (offset >= length) { |
50 | return string_t((uint32_t)0); |
51 | } |
52 | return SubstringFun::substring_ascii_only(result, input_data, offset, length - offset); |
53 | } |
54 | |
55 | // size is at most the input size: alloc it |
56 | idx_t required_len = input_size + 1; |
57 | if (required_len > current_len) { |
58 | // need a resize |
59 | current_len = required_len; |
60 | output = unique_ptr<char[]>{new char[required_len]}; |
61 | } |
62 | |
63 | // use grapheme iterator to iterate over the characters |
64 | int current_offset = 0; |
65 | int output_size = 0; |
66 | utf8proc_grapheme_callback(input_data, input_size, [&](size_t start, size_t end) { |
67 | if (current_offset >= offset) { |
68 | // this character belongs to the output: copy it there |
69 | memcpy(output.get() + output_size, input_data + start, end - start); |
70 | output_size += end - start; |
71 | } |
72 | current_offset++; |
73 | // stop iterating after we have exceeded the required characters |
74 | return current_offset < offset + length; |
75 | }); |
76 | output[output_size] = '\0'; |
77 | return StringVector::AddString(result, output.get(), output_size); |
78 | } |
79 | |
80 | static void substring_function(DataChunk &args, ExpressionState &state, Vector &result) { |
81 | assert(args.column_count() == 3 && args.data[0].type == TypeId::VARCHAR && args.data[1].type == TypeId::INT32 && |
82 | args.data[2].type == TypeId::INT32); |
83 | auto &input_vector = args.data[0]; |
84 | auto &offset_vector = args.data[1]; |
85 | auto &length_vector = args.data[2]; |
86 | |
87 | idx_t current_len = 0; |
88 | unique_ptr<char[]> output; |
89 | TernaryExecutor::Execute<string_t, int, int, string_t>( |
90 | input_vector, offset_vector, length_vector, result, args.size(), |
91 | [&](string_t input_string, int offset, int length) { |
92 | return SubstringFun::substring_scalar_function(result, input_string, offset, length, output, current_len); |
93 | }); |
94 | } |
95 | |
96 | void SubstringFun::RegisterFunction(BuiltinFunctions &set) { |
97 | set.AddFunction({"substring" , "substr" }, ScalarFunction({SQLType::VARCHAR, SQLType::INTEGER, SQLType::INTEGER}, |
98 | SQLType::VARCHAR, substring_function)); |
99 | } |
100 | |
101 | } // namespace duckdb |
102 | |