1 | /*------------------------------------------------------------------------- |
2 | * |
3 | * scansup.c |
4 | * support routines for the lex/flex scanner, used by both the normal |
5 | * backend as well as the bootstrap backend |
6 | * |
7 | * Portions Copyright (c) 1996-2019, PostgreSQL Global Development Group |
8 | * Portions Copyright (c) 1994, Regents of the University of California |
9 | * |
10 | * |
11 | * IDENTIFICATION |
12 | * src/backend/parser/scansup.c |
13 | * |
14 | *------------------------------------------------------------------------- |
15 | */ |
16 | #include "postgres.h" |
17 | |
18 | #include <ctype.h> |
19 | |
20 | #include "parser/scansup.h" |
21 | #include "mb/pg_wchar.h" |
22 | |
23 | |
24 | /* ---------------- |
25 | * scanstr |
26 | * |
27 | * if the string passed in has escaped codes, map the escape codes to actual |
28 | * chars |
29 | * |
30 | * the string returned is palloc'd and should eventually be pfree'd by the |
31 | * caller! |
32 | * ---------------- |
33 | */ |
34 | |
35 | char * |
36 | scanstr(const char *s) |
37 | { |
38 | char *newStr; |
39 | int len, |
40 | i, |
41 | j; |
42 | |
43 | if (s == NULL || s[0] == '\0') |
44 | return pstrdup("" ); |
45 | |
46 | len = strlen(s); |
47 | |
48 | newStr = palloc(len + 1); /* string cannot get longer */ |
49 | |
50 | for (i = 0, j = 0; i < len; i++) |
51 | { |
52 | if (s[i] == '\'') |
53 | { |
54 | /* |
55 | * Note: if scanner is working right, unescaped quotes can only |
56 | * appear in pairs, so there should be another character. |
57 | */ |
58 | i++; |
59 | /* The bootstrap parser is not as smart, so check here. */ |
60 | Assert(s[i] == '\''); |
61 | newStr[j] = s[i]; |
62 | } |
63 | else if (s[i] == '\\') |
64 | { |
65 | i++; |
66 | switch (s[i]) |
67 | { |
68 | case 'b': |
69 | newStr[j] = '\b'; |
70 | break; |
71 | case 'f': |
72 | newStr[j] = '\f'; |
73 | break; |
74 | case 'n': |
75 | newStr[j] = '\n'; |
76 | break; |
77 | case 'r': |
78 | newStr[j] = '\r'; |
79 | break; |
80 | case 't': |
81 | newStr[j] = '\t'; |
82 | break; |
83 | case '0': |
84 | case '1': |
85 | case '2': |
86 | case '3': |
87 | case '4': |
88 | case '5': |
89 | case '6': |
90 | case '7': |
91 | { |
92 | int k; |
93 | long octVal = 0; |
94 | |
95 | for (k = 0; |
96 | s[i + k] >= '0' && s[i + k] <= '7' && k < 3; |
97 | k++) |
98 | octVal = (octVal << 3) + (s[i + k] - '0'); |
99 | i += k - 1; |
100 | newStr[j] = ((char) octVal); |
101 | } |
102 | break; |
103 | default: |
104 | newStr[j] = s[i]; |
105 | break; |
106 | } /* switch */ |
107 | } /* s[i] == '\\' */ |
108 | else |
109 | newStr[j] = s[i]; |
110 | j++; |
111 | } |
112 | newStr[j] = '\0'; |
113 | return newStr; |
114 | } |
115 | |
116 | |
117 | /* |
118 | * downcase_truncate_identifier() --- do appropriate downcasing and |
119 | * truncation of an unquoted identifier. Optionally warn of truncation. |
120 | * |
121 | * Returns a palloc'd string containing the adjusted identifier. |
122 | * |
123 | * Note: in some usages the passed string is not null-terminated. |
124 | * |
125 | * Note: the API of this function is designed to allow for downcasing |
126 | * transformations that increase the string length, but we don't yet |
127 | * support that. If you want to implement it, you'll need to fix |
128 | * SplitIdentifierString() in utils/adt/varlena.c. |
129 | */ |
130 | char * |
131 | downcase_truncate_identifier(const char *ident, int len, bool warn) |
132 | { |
133 | return downcase_identifier(ident, len, warn, true); |
134 | } |
135 | |
136 | /* |
137 | * a workhorse for downcase_truncate_identifier |
138 | */ |
139 | char * |
140 | downcase_identifier(const char *ident, int len, bool warn, bool truncate) |
141 | { |
142 | char *result; |
143 | int i; |
144 | bool enc_is_single_byte; |
145 | |
146 | result = palloc(len + 1); |
147 | enc_is_single_byte = pg_database_encoding_max_length() == 1; |
148 | |
149 | /* |
150 | * SQL99 specifies Unicode-aware case normalization, which we don't yet |
151 | * have the infrastructure for. Instead we use tolower() to provide a |
152 | * locale-aware translation. However, there are some locales where this |
153 | * is not right either (eg, Turkish may do strange things with 'i' and |
154 | * 'I'). Our current compromise is to use tolower() for characters with |
155 | * the high bit set, as long as they aren't part of a multi-byte |
156 | * character, and use an ASCII-only downcasing for 7-bit characters. |
157 | */ |
158 | for (i = 0; i < len; i++) |
159 | { |
160 | unsigned char ch = (unsigned char) ident[i]; |
161 | |
162 | if (ch >= 'A' && ch <= 'Z') |
163 | ch += 'a' - 'A'; |
164 | else if (enc_is_single_byte && IS_HIGHBIT_SET(ch) && isupper(ch)) |
165 | ch = tolower(ch); |
166 | result[i] = (char) ch; |
167 | } |
168 | result[i] = '\0'; |
169 | |
170 | if (i >= NAMEDATALEN && truncate) |
171 | truncate_identifier(result, i, warn); |
172 | |
173 | return result; |
174 | } |
175 | |
176 | |
177 | /* |
178 | * truncate_identifier() --- truncate an identifier to NAMEDATALEN-1 bytes. |
179 | * |
180 | * The given string is modified in-place, if necessary. A warning is |
181 | * issued if requested. |
182 | * |
183 | * We require the caller to pass in the string length since this saves a |
184 | * strlen() call in some common usages. |
185 | */ |
186 | void |
187 | truncate_identifier(char *ident, int len, bool warn) |
188 | { |
189 | if (len >= NAMEDATALEN) |
190 | { |
191 | len = pg_mbcliplen(ident, len, NAMEDATALEN - 1); |
192 | if (warn) |
193 | { |
194 | /* |
195 | * We avoid using %.*s here because it can misbehave if the data |
196 | * is not valid in what libc thinks is the prevailing encoding. |
197 | */ |
198 | char buf[NAMEDATALEN]; |
199 | |
200 | memcpy(buf, ident, len); |
201 | buf[len] = '\0'; |
202 | ereport(NOTICE, |
203 | (errcode(ERRCODE_NAME_TOO_LONG), |
204 | errmsg("identifier \"%s\" will be truncated to \"%s\"" , |
205 | ident, buf))); |
206 | } |
207 | ident[len] = '\0'; |
208 | } |
209 | } |
210 | |
211 | /* |
212 | * scanner_isspace() --- return true if flex scanner considers char whitespace |
213 | * |
214 | * This should be used instead of the potentially locale-dependent isspace() |
215 | * function when it's important to match the lexer's behavior. |
216 | * |
217 | * In principle we might need similar functions for isalnum etc, but for the |
218 | * moment only isspace seems needed. |
219 | */ |
220 | bool |
221 | scanner_isspace(char ch) |
222 | { |
223 | /* This must match scan.l's list of {space} characters */ |
224 | if (ch == ' ' || |
225 | ch == '\t' || |
226 | ch == '\n' || |
227 | ch == '\r' || |
228 | ch == '\f') |
229 | return true; |
230 | return false; |
231 | } |
232 | |