Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 1 | .TH PCRE2SERIALIZE 3 "27 June 2018" "PCRE2 10.32" |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 2 | .SH NAME |
| 3 | PCRE2 - Perl-compatible regular expressions (revised API) |
| 4 | .SH "SAVING AND RE-USING PRECOMPILED PCRE2 PATTERNS" |
| 5 | .rs |
| 6 | .sp |
| 7 | .nf |
| 8 | .B int32_t pcre2_serialize_decode(pcre2_code **\fIcodes\fP, |
Elliott Hughes | 4e19c8e | 2022-04-15 15:11:02 -0700 | [diff] [blame] | 9 | .B " int32_t \fInumber_of_codes\fP, const uint8_t *\fIbytes\fP," |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 10 | .B " pcre2_general_context *\fIgcontext\fP);" |
| 11 | .sp |
Elliott Hughes | 4e19c8e | 2022-04-15 15:11:02 -0700 | [diff] [blame] | 12 | .B int32_t pcre2_serialize_encode(const pcre2_code **\fIcodes\fP, |
| 13 | .B " int32_t \fInumber_of_codes\fP, uint8_t **\fIserialized_bytes\fP," |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 14 | .B " PCRE2_SIZE *\fIserialized_size\fP, pcre2_general_context *\fIgcontext\fP);" |
| 15 | .sp |
| 16 | .B void pcre2_serialize_free(uint8_t *\fIbytes\fP); |
| 17 | .sp |
| 18 | .B int32_t pcre2_serialize_get_number_of_codes(const uint8_t *\fIbytes\fP); |
| 19 | .fi |
| 20 | .sp |
| 21 | If you are running an application that uses a large number of regular |
| 22 | expression patterns, it may be useful to store them in a precompiled form |
| 23 | instead of having to compile them every time the application is run. However, |
| 24 | if you are using the just-in-time optimization feature, it is not possible to |
| 25 | save and reload the JIT data, because it is position-dependent. The host on |
| 26 | which the patterns are reloaded must be running the same version of PCRE2, with |
| 27 | the same code unit width, and must also have the same endianness, pointer width |
| 28 | and PCRE2_SIZE type. For example, patterns compiled on a 32-bit system using |
| 29 | PCRE2's 16-bit library cannot be reloaded on a 64-bit system, nor can they be |
| 30 | reloaded using the 8-bit library. |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 31 | .P |
| 32 | Note that "serialization" in PCRE2 does not convert compiled patterns to an |
| 33 | abstract format like Java or .NET serialization. The serialized output is |
| 34 | really just a bytecode dump, which is why it can only be reloaded in the same |
| 35 | environment as the one that created it. Hence the restrictions mentioned above. |
| 36 | Applications that are not statically linked with a fixed version of PCRE2 must |
| 37 | be prepared to recompile patterns from their sources, in order to be immune to |
| 38 | PCRE2 upgrades. |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 39 | . |
| 40 | . |
Janis Danisevskis | 8b979b2 | 2016-08-15 16:09:16 +0100 | [diff] [blame] | 41 | .SH "SECURITY CONCERNS" |
| 42 | .rs |
| 43 | .sp |
| 44 | The facility for saving and restoring compiled patterns is intended for use |
| 45 | within individual applications. As such, the data supplied to |
| 46 | \fBpcre2_serialize_decode()\fP is expected to be trusted data, not data from |
| 47 | arbitrary external sources. There is only some simple consistency checking, not |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 48 | complete validation of what is being re-loaded. Corrupted data may cause |
| 49 | undefined results. For example, if the length field of a pattern in the |
| 50 | serialized data is corrupted, the deserializing code may read beyond the end of |
| 51 | the byte stream that is passed to it. |
Janis Danisevskis | 8b979b2 | 2016-08-15 16:09:16 +0100 | [diff] [blame] | 52 | . |
| 53 | . |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 54 | .SH "SAVING COMPILED PATTERNS" |
| 55 | .rs |
| 56 | .sp |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 57 | Before compiled patterns can be saved they must be serialized, which in PCRE2 |
| 58 | means converting the pattern to a stream of bytes. A single byte stream may |
| 59 | contain any number of compiled patterns, but they must all use the same |
| 60 | character tables. A single copy of the tables is included in the byte stream |
| 61 | (its size is 1088 bytes). For more details of character tables, see the |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 62 | .\" HTML <a href="pcre2api.html#localesupport"> |
| 63 | .\" </a> |
| 64 | section on locale support |
| 65 | .\" |
| 66 | in the |
| 67 | .\" HREF |
| 68 | \fBpcre2api\fP |
| 69 | .\" |
| 70 | documentation. |
| 71 | .P |
| 72 | The function \fBpcre2_serialize_encode()\fP creates a serialized byte stream |
| 73 | from a list of compiled patterns. Its first two arguments specify the list, |
| 74 | being a pointer to a vector of pointers to compiled patterns, and the length of |
| 75 | the vector. The third and fourth arguments point to variables which are set to |
| 76 | point to the created byte stream and its length, respectively. The final |
| 77 | argument is a pointer to a general context, which can be used to specify custom |
| 78 | memory mangagement functions. If this argument is NULL, \fBmalloc()\fP is used |
| 79 | to obtain memory for the byte stream. The yield of the function is the number |
| 80 | of serialized patterns, or one of the following negative error codes: |
| 81 | .sp |
| 82 | PCRE2_ERROR_BADDATA the number of patterns is zero or less |
| 83 | PCRE2_ERROR_BADMAGIC mismatch of id bytes in one of the patterns |
| 84 | PCRE2_ERROR_MEMORY memory allocation failed |
| 85 | PCRE2_ERROR_MIXEDTABLES the patterns do not all use the same tables |
| 86 | PCRE2_ERROR_NULL the 1st, 3rd, or 4th argument is NULL |
| 87 | .sp |
| 88 | PCRE2_ERROR_BADMAGIC means either that a pattern's code has been corrupted, or |
| 89 | that a slot in the vector does not point to a compiled pattern. |
| 90 | .P |
| 91 | Once a set of patterns has been serialized you can save the data in any |
| 92 | appropriate manner. Here is sample code that compiles two patterns and writes |
| 93 | them to a file. It assumes that the variable \fIfd\fP refers to a file that is |
| 94 | open for output. The error checking that should be present in a real |
| 95 | application has been omitted for simplicity. |
| 96 | .sp |
| 97 | int errorcode; |
| 98 | uint8_t *bytes; |
| 99 | PCRE2_SIZE erroroffset; |
| 100 | PCRE2_SIZE bytescount; |
| 101 | pcre2_code *list_of_codes[2]; |
| 102 | list_of_codes[0] = pcre2_compile("first pattern", |
| 103 | PCRE2_ZERO_TERMINATED, 0, &errorcode, &erroroffset, NULL); |
| 104 | list_of_codes[1] = pcre2_compile("second pattern", |
| 105 | PCRE2_ZERO_TERMINATED, 0, &errorcode, &erroroffset, NULL); |
| 106 | errorcode = pcre2_serialize_encode(list_of_codes, 2, &bytes, |
| 107 | &bytescount, NULL); |
| 108 | errorcode = fwrite(bytes, 1, bytescount, fd); |
| 109 | .sp |
| 110 | Note that the serialized data is binary data that may contain any of the 256 |
| 111 | possible byte values. On systems that make a distinction between binary and |
| 112 | non-binary data, be sure that the file is opened for binary output. |
| 113 | .P |
| 114 | Serializing a set of patterns leaves the original data untouched, so they can |
| 115 | still be used for matching. Their memory must eventually be freed in the usual |
| 116 | way by calling \fBpcre2_code_free()\fP. When you have finished with the byte |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 117 | stream, it too must be freed by calling \fBpcre2_serialize_free()\fP. If this |
| 118 | function is called with a NULL argument, it returns immediately without doing |
| 119 | anything. |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 120 | . |
| 121 | . |
| 122 | .SH "RE-USING PRECOMPILED PATTERNS" |
| 123 | .rs |
| 124 | .sp |
| 125 | In order to re-use a set of saved patterns you must first make the serialized |
| 126 | byte stream available in main memory (for example, by reading from a file). The |
| 127 | management of this memory block is up to the application. You can use the |
| 128 | \fBpcre2_serialize_get_number_of_codes()\fP function to find out how many |
| 129 | compiled patterns are in the serialized data without actually decoding the |
| 130 | patterns: |
| 131 | .sp |
| 132 | uint8_t *bytes = <serialized data>; |
| 133 | int32_t number_of_codes = pcre2_serialize_get_number_of_codes(bytes); |
| 134 | .sp |
| 135 | The \fBpcre2_serialize_decode()\fP function reads a byte stream and recreates |
| 136 | the compiled patterns in new memory blocks, setting pointers to them in a |
| 137 | vector. The first two arguments are a pointer to a suitable vector and its |
| 138 | length, and the third argument points to a byte stream. The final argument is a |
| 139 | pointer to a general context, which can be used to specify custom memory |
| 140 | mangagement functions for the decoded patterns. If this argument is NULL, |
| 141 | \fBmalloc()\fP and \fBfree()\fP are used. After deserialization, the byte |
| 142 | stream is no longer needed and can be discarded. |
| 143 | .sp |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 144 | pcre2_code *list_of_codes[2]; |
| 145 | uint8_t *bytes = <serialized data>; |
| 146 | int32_t number_of_codes = |
| 147 | pcre2_serialize_decode(list_of_codes, 2, bytes, NULL); |
| 148 | .sp |
| 149 | If the vector is not large enough for all the patterns in the byte stream, it |
| 150 | is filled with those that fit, and the remainder are ignored. The yield of the |
| 151 | function is the number of decoded patterns, or one of the following negative |
| 152 | error codes: |
| 153 | .sp |
Janis Danisevskis | 8b979b2 | 2016-08-15 16:09:16 +0100 | [diff] [blame] | 154 | PCRE2_ERROR_BADDATA second argument is zero or less |
| 155 | PCRE2_ERROR_BADMAGIC mismatch of id bytes in the data |
| 156 | PCRE2_ERROR_BADMODE mismatch of code unit size or PCRE2 version |
| 157 | PCRE2_ERROR_BADSERIALIZEDDATA other sanity check failure |
| 158 | PCRE2_ERROR_MEMORY memory allocation failed |
| 159 | PCRE2_ERROR_NULL first or third argument is NULL |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 160 | .sp |
| 161 | PCRE2_ERROR_BADMAGIC may mean that the data is corrupt, or that it was compiled |
| 162 | on a system with different endianness. |
| 163 | .P |
| 164 | Decoded patterns can be used for matching in the usual way, and must be freed |
| 165 | by calling \fBpcre2_code_free()\fP. However, be aware that there is a potential |
| 166 | race issue if you are using multiple patterns that were decoded from a single |
| 167 | byte stream in a multithreaded application. A single copy of the character |
| 168 | tables is used by all the decoded patterns and a reference count is used to |
| 169 | arrange for its memory to be automatically freed when the last pattern is |
| 170 | freed, but there is no locking on this reference count. Therefore, if you want |
| 171 | to call \fBpcre2_code_free()\fP for these patterns in different threads, you |
| 172 | must arrange your own locking, and ensure that \fBpcre2_code_free()\fP cannot |
| 173 | be called by two threads at the same time. |
| 174 | .P |
| 175 | If a pattern was processed by \fBpcre2_jit_compile()\fP before being |
| 176 | serialized, the JIT data is discarded and so is no longer available after a |
| 177 | save/restore cycle. You can, however, process a restored pattern with |
| 178 | \fBpcre2_jit_compile()\fP if you wish. |
| 179 | . |
| 180 | . |
| 181 | . |
| 182 | .SH AUTHOR |
| 183 | .rs |
| 184 | .sp |
| 185 | .nf |
| 186 | Philip Hazel |
| 187 | University Computing Service |
| 188 | Cambridge, England. |
| 189 | .fi |
| 190 | . |
| 191 | . |
| 192 | .SH REVISION |
| 193 | .rs |
| 194 | .sp |
| 195 | .nf |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 196 | Last updated: 27 June 2018 |
| 197 | Copyright (c) 1997-2018 University of Cambridge. |
Janis Danisevskis | 53e448c | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 198 | .fi |