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) 1996-2020, 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 "mb/pg_wchar.h"
21 #include "parser/scansup.h"
22
23 /* ----------------
24 * scanstr
25 *
26 * if the string passed in has escaped codes, map the escape codes to actual
27 * chars
28 *
29 * the string returned is palloc'd and should eventually be pfree'd by the
30 * caller!
31 * ----------------
32 */
33
34 char *
scanstr(const char * s)35 scanstr(const char *s)
36 {
37 char *newStr;
38 int len,
39 i,
40 j;
41
42 if (s == NULL || s[0] == '\0')
43 return pstrdup("");
44
45 len = strlen(s);
46
47 newStr = palloc(len + 1); /* string cannot get longer */
48
49 for (i = 0, j = 0; i < len; i++)
50 {
51 if (s[i] == '\'')
52 {
53 /*
54 * Note: if scanner is working right, unescaped quotes can only
55 * appear in pairs, so there should be another character.
56 */
57 i++;
58 /* The bootstrap parser is not as smart, so check here. */
59 Assert(s[i] == '\'');
60 newStr[j] = s[i];
61 }
62 else if (s[i] == '\\')
63 {
64 i++;
65 switch (s[i])
66 {
67 case 'b':
68 newStr[j] = '\b';
69 break;
70 case 'f':
71 newStr[j] = '\f';
72 break;
73 case 'n':
74 newStr[j] = '\n';
75 break;
76 case 'r':
77 newStr[j] = '\r';
78 break;
79 case 't':
80 newStr[j] = '\t';
81 break;
82 case '0':
83 case '1':
84 case '2':
85 case '3':
86 case '4':
87 case '5':
88 case '6':
89 case '7':
90 {
91 int k;
92 long octVal = 0;
93
94 for (k = 0;
95 s[i + k] >= '0' && s[i + k] <= '7' && k < 3;
96 k++)
97 octVal = (octVal << 3) + (s[i + k] - '0');
98 i += k - 1;
99 newStr[j] = ((char) octVal);
100 }
101 break;
102 default:
103 newStr[j] = s[i];
104 break;
105 } /* switch */
106 } /* s[i] == '\\' */
107 else
108 newStr[j] = s[i];
109 j++;
110 }
111 newStr[j] = '\0';
112 return newStr;
113 }
114
115
116 /*
117 * downcase_truncate_identifier() --- do appropriate downcasing and
118 * truncation of an unquoted identifier. Optionally warn of truncation.
119 *
120 * Returns a palloc'd string containing the adjusted identifier.
121 *
122 * Note: in some usages the passed string is not null-terminated.
123 *
124 * Note: the API of this function is designed to allow for downcasing
125 * transformations that increase the string length, but we don't yet
126 * support that. If you want to implement it, you'll need to fix
127 * SplitIdentifierString() in utils/adt/varlena.c.
128 */
129 char *
downcase_truncate_identifier(const char * ident,int len,bool warn)130 downcase_truncate_identifier(const char *ident, int len, bool warn)
131 {
132 return downcase_identifier(ident, len, warn, true);
133 }
134
135 /*
136 * a workhorse for downcase_truncate_identifier
137 */
138 char *
downcase_identifier(const char * ident,int len,bool warn,bool truncate)139 downcase_identifier(const char *ident, int len, bool warn, bool truncate)
140 {
141 char *result;
142 int i;
143 bool enc_is_single_byte;
144
145 result = palloc(len + 1);
146 enc_is_single_byte = pg_database_encoding_max_length() == 1;
147
148 /*
149 * SQL99 specifies Unicode-aware case normalization, which we don't yet
150 * have the infrastructure for. Instead we use tolower() to provide a
151 * locale-aware translation. However, there are some locales where this
152 * is not right either (eg, Turkish may do strange things with 'i' and
153 * 'I'). Our current compromise is to use tolower() for characters with
154 * the high bit set, as long as they aren't part of a multi-byte
155 * character, and use an ASCII-only downcasing for 7-bit characters.
156 */
157 for (i = 0; i < len; i++)
158 {
159 unsigned char ch = (unsigned char) ident[i];
160
161 if (ch >= 'A' && ch <= 'Z')
162 ch += 'a' - 'A';
163 else if (enc_is_single_byte && IS_HIGHBIT_SET(ch) && isupper(ch))
164 ch = tolower(ch);
165 result[i] = (char) ch;
166 }
167 result[i] = '\0';
168
169 if (i >= NAMEDATALEN && truncate)
170 truncate_identifier(result, i, warn);
171
172 return result;
173 }
174
175
176 /*
177 * truncate_identifier() --- truncate an identifier to NAMEDATALEN-1 bytes.
178 *
179 * The given string is modified in-place, if necessary. A warning is
180 * issued if requested.
181 *
182 * We require the caller to pass in the string length since this saves a
183 * strlen() call in some common usages.
184 */
185 void
truncate_identifier(char * ident,int len,bool warn)186 truncate_identifier(char *ident, int len, bool warn)
187 {
188 if (len >= NAMEDATALEN)
189 {
190 len = pg_mbcliplen(ident, len, NAMEDATALEN - 1);
191 if (warn)
192 {
193 /*
194 * We avoid using %.*s here because it can misbehave if the data
195 * is not valid in what libc thinks is the prevailing encoding.
196 */
197 char buf[NAMEDATALEN];
198
199 memcpy(buf, ident, len);
200 buf[len] = '\0';
201 ereport(NOTICE,
202 (errcode(ERRCODE_NAME_TOO_LONG),
203 errmsg("identifier \"%s\" will be truncated to \"%s\"",
204 ident, buf)));
205 }
206 ident[len] = '\0';
207 }
208 }
209
210 /*
211 * scanner_isspace() --- return true if flex scanner considers char whitespace
212 *
213 * This should be used instead of the potentially locale-dependent isspace()
214 * function when it's important to match the lexer's behavior.
215 *
216 * In principle we might need similar functions for isalnum etc, but for the
217 * moment only isspace seems needed.
218 */
219 bool
scanner_isspace(char ch)220 scanner_isspace(char ch)
221 {
222 /* This must match scan.l's list of {space} characters */
223 if (ch == ' ' ||
224 ch == '\t' ||
225 ch == '\n' ||
226 ch == '\r' ||
227 ch == '\f')
228 return true;
229 return false;
230 }
231