varmod-match.mk revision 1.18 1 1.18 rillig # $NetBSD: varmod-match.mk,v 1.18 2023/12/16 23:38:35 rillig Exp $
2 1.1 rillig #
3 1.18 rillig # Tests for the ':M' modifier, which keeps only those words that match the
4 1.2 rillig # given pattern.
5 1.4 rillig #
6 1.18 rillig # Table of contents
7 1.18 rillig #
8 1.18 rillig # 1. Pattern characters '*', '?' and '\'
9 1.18 rillig # 2. Character lists and character ranges
10 1.18 rillig # 3. Parsing and escaping
11 1.18 rillig # 4. Interaction with other modifiers
12 1.18 rillig # 5. Performance
13 1.18 rillig # 6. Error handling
14 1.18 rillig # 7. Historical bugs
15 1.3 rillig
16 1.3 rillig NUMBERS= One Two Three Four five six seven
17 1.3 rillig
18 1.18 rillig # 1. Pattern characters '*', '?' and '\'
19 1.5 rillig #
20 1.18 rillig # * matches 0 or more characters
21 1.18 rillig # ? matches 1 character
22 1.18 rillig # \x matches the character 'x'
23 1.3 rillig
24 1.18 rillig # TODO
25 1.4 rillig
26 1.10 rillig
27 1.18 rillig # 2. Character lists and character ranges
28 1.5 rillig #
29 1.18 rillig # [...] matches 1 character from the listed characters
30 1.18 rillig # [^...] matches 1 character from the unlisted characters
31 1.18 rillig # [a-z] matches 1 character from the range 'a' to 'z'
32 1.18 rillig # [z-a] matches 1 character from the range 'a' to 'z'
33 1.18 rillig
34 1.18 rillig # Only keep words that start with an uppercase letter.
35 1.18 rillig .if ${NUMBERS:M[A-Z]*} != "One Two Three Four"
36 1.4 rillig . error
37 1.4 rillig .endif
38 1.4 rillig
39 1.18 rillig # Only keep words that start with a character other than an uppercase letter.
40 1.18 rillig .if ${NUMBERS:M[^A-Z]*} != "five six seven"
41 1.4 rillig . error
42 1.4 rillig .endif
43 1.4 rillig
44 1.7 rillig # [] matches never
45 1.7 rillig .if ${ ab a[]b a[b a b :L:M[]} != ""
46 1.7 rillig . error
47 1.7 rillig .endif
48 1.7 rillig
49 1.7 rillig # a[]b matches never
50 1.7 rillig .if ${ ab a[]b a[b a b [ ] :L:Ma[]b} != ""
51 1.7 rillig . error
52 1.7 rillig .endif
53 1.7 rillig
54 1.7 rillig # [^] matches exactly 1 arbitrary character
55 1.7 rillig .if ${ ab a[]b a[b a b [ ] :L:M[^]} != "a b [ ]"
56 1.7 rillig . error
57 1.7 rillig .endif
58 1.7 rillig
59 1.7 rillig # a[^]b matches 'a', then exactly 1 arbitrary character, then 'b'
60 1.7 rillig .if ${ ab a[]b a[b a b :L:Ma[^]b} != "a[b"
61 1.7 rillig . error
62 1.7 rillig .endif
63 1.7 rillig
64 1.7 rillig # [Nn0] matches exactly 1 character from the set 'N', 'n', '0'
65 1.7 rillig .if ${ a b N n 0 Nn0 [ ] :L:M[Nn0]} != "N n 0"
66 1.7 rillig . error
67 1.7 rillig .endif
68 1.7 rillig
69 1.7 rillig # [a-c] matches exactly 1 character from the range 'a' to 'c'
70 1.7 rillig .if ${ A B C a b c d [a-c] [a] :L:M[a-c]} != "a b c"
71 1.7 rillig . error
72 1.7 rillig .endif
73 1.7 rillig
74 1.7 rillig # [c-a] matches the same as [a-c]
75 1.7 rillig .if ${ A B C a b c d [a-c] [a] :L:M[c-a]} != "a b c"
76 1.7 rillig . error
77 1.7 rillig .endif
78 1.7 rillig
79 1.7 rillig # [^a-c67]
80 1.8 rillig # matches a single character, except for 'a', 'b', 'c', '6' or
81 1.8 rillig # '7'
82 1.7 rillig .if ${ A B C a b c d 5 6 7 8 [a-c] [a] :L:M[^a-c67]} != "A B C d 5 8"
83 1.7 rillig . error
84 1.7 rillig .endif
85 1.7 rillig
86 1.18 rillig # [\] matches a single backslash; no escaping takes place in
87 1.18 rillig # character ranges
88 1.18 rillig # Without the 'b' in the below words, the backslash would end a word and thus
89 1.18 rillig # influence how the string is split into words.
90 1.18 rillig WORDS= a\b a[\]b ab a\\b
91 1.11 rillig .if ${WORDS:Ma[\]b} != "a\\b"
92 1.11 rillig . error
93 1.11 rillig .endif
94 1.11 rillig
95 1.18 rillig # [[-]] May look like it would match a single '[', '\' or ']', but
96 1.18 rillig # the inner ']' has two roles: it is the upper bound of the
97 1.18 rillig # character range as well as the closing character of the
98 1.18 rillig # character list. The outer ']' is just a regular character.
99 1.18 rillig WORDS= [ ] [] \] ]]
100 1.18 rillig .if ${WORDS:M[[-]]} != "[] \\] ]]"
101 1.18 rillig . error
102 1.18 rillig .endif
103 1.18 rillig
104 1.18 rillig # [b[-]a]
105 1.18 rillig # Same as for '[[-]]': the character list stops at the first
106 1.18 rillig # ']', and the 'a]' is treated as a literal string.
107 1.18 rillig WORDS= [a \a ]a []a \]a ]]a [a] \a] ]a] ba]
108 1.18 rillig .if ${WORDS:M[b[-]a]} != "[a] \\a] ]a] ba]"
109 1.18 rillig . error
110 1.18 rillig .endif
111 1.18 rillig
112 1.18 rillig # [-] Matches a single '-' since the '-' only becomes part of a
113 1.18 rillig # character range if it is preceded and followed by another
114 1.18 rillig # character.
115 1.18 rillig WORDS= - -]
116 1.18 rillig .if ${WORDS:M[-]} != "-"
117 1.18 rillig . error
118 1.18 rillig .endif
119 1.18 rillig
120 1.18 rillig # Only keep words that don't start with s and at the same time end with
121 1.18 rillig # either of [ex].
122 1.18 rillig #
123 1.18 rillig # This test case ensures that the negation from the first character class
124 1.18 rillig # does not propagate to the second character class.
125 1.18 rillig .if ${NUMBERS:M[^s]*[ex]} != "One Three five"
126 1.18 rillig . error
127 1.18 rillig .endif
128 1.18 rillig
129 1.18 rillig
130 1.18 rillig # 3. Parsing and escaping
131 1.18 rillig #
132 1.18 rillig # * matches 0 or more characters
133 1.18 rillig # ? matches 1 character
134 1.18 rillig # \ outside a character list, escapes the following character
135 1.18 rillig # [ starts a character list for matching 1 character
136 1.18 rillig # ] ends a character list for matching 1 character
137 1.18 rillig # - in a character list, forms a character range
138 1.18 rillig # ^ at the beginning of a character list, negates the list
139 1.18 rillig # ( while parsing the pattern, starts a nesting level
140 1.18 rillig # ) while parsing the pattern, ends a nesting level
141 1.18 rillig # { while parsing the pattern, starts a nesting level
142 1.18 rillig # } while parsing the pattern, ends a nesting level
143 1.18 rillig # : while parsing the pattern, terminates the pattern
144 1.18 rillig # $ while parsing the pattern, starts a nested expression
145 1.18 rillig # # in a line except a shell command, starts a comment
146 1.18 rillig
147 1.18 rillig # The pattern can come from an expression. For single-letter
148 1.18 rillig # variables, either the short form or the long form can be used, just as
149 1.18 rillig # everywhere else.
150 1.18 rillig PRIMES= 2 3 5 7 11
151 1.18 rillig n= 2
152 1.18 rillig .if ${PRIMES:M$n} != "2"
153 1.18 rillig . error
154 1.18 rillig .endif
155 1.18 rillig .if ${PRIMES:M${n}} != "2"
156 1.18 rillig . error
157 1.18 rillig .endif
158 1.18 rillig .if ${PRIMES:M${:U2}} != "2"
159 1.18 rillig . error
160 1.18 rillig .endif
161 1.18 rillig
162 1.7 rillig # : terminates the pattern
163 1.7 rillig .if ${ A * :L:M:} != ""
164 1.7 rillig . error
165 1.7 rillig .endif
166 1.7 rillig
167 1.7 rillig # \: matches a colon
168 1.7 rillig .if ${ ${:U\: \:\:} :L:M\:} != ":"
169 1.7 rillig . error
170 1.7 rillig .endif
171 1.7 rillig
172 1.7 rillig # ${:U\:} matches a colon
173 1.7 rillig .if ${ ${:U\:} ${:U\:\:} :L:M${:U\:}} != ":"
174 1.7 rillig . error
175 1.7 rillig .endif
176 1.7 rillig
177 1.18 rillig # To match a dollar sign in a word, double it.
178 1.18 rillig #
179 1.18 rillig # This is different from the :S and :C variable modifiers, where a '$'
180 1.18 rillig # has to be escaped as '\$'.
181 1.18 rillig .if ${:Ua \$ sign:M*$$*} != "\$"
182 1.7 rillig . error
183 1.18 rillig .endif
184 1.18 rillig
185 1.18 rillig # In the :M modifier, '\$' does not escape a dollar. Instead it is
186 1.18 rillig # interpreted as a backslash followed by whatever expression the
187 1.18 rillig # '$' starts.
188 1.18 rillig #
189 1.18 rillig # This differs from the :S, :C and several other variable modifiers.
190 1.18 rillig ${:U*}= asterisk
191 1.18 rillig .if ${:Ua \$ sign any-asterisk:M*\$*} != "any-asterisk"
192 1.7 rillig . error
193 1.7 rillig .endif
194 1.7 rillig
195 1.18 rillig # TODO: ${VAR:M(((}}}}
196 1.18 rillig # TODO: ${VAR:M{{{)))}
197 1.18 rillig # TODO: ${VAR:M${UNBALANCED}}
198 1.18 rillig # TODO: ${VAR:M${:U(((\}\}\}}}
199 1.18 rillig
200 1.18 rillig
201 1.18 rillig # 4. Interaction with other modifiers
202 1.18 rillig
203 1.18 rillig # The modifier ':tW' prevents splitting at whitespace. Even leading and
204 1.18 rillig # trailing whitespace is preserved.
205 1.18 rillig .if ${ plain string :L:tW:M*} != " plain string "
206 1.11 rillig . error
207 1.11 rillig .endif
208 1.11 rillig
209 1.18 rillig # Without the modifier ':tW', the string is split into words. All whitespace
210 1.18 rillig # around and between the words is normalized to a single space.
211 1.18 rillig .if ${ plain string :L:M*} != "plain string"
212 1.11 rillig . error
213 1.11 rillig .endif
214 1.11 rillig
215 1.18 rillig
216 1.18 rillig # 5. Performance
217 1.18 rillig
218 1.18 rillig # Before 2020-06-13, this expression called Str_Match 601,080,390 times.
219 1.18 rillig # Since 2020-06-13, this expression calls Str_Match 1 time.
220 1.18 rillig .if ${:U****************:M****************b}
221 1.11 rillig .endif
222 1.11 rillig
223 1.18 rillig # Before 2023-06-22, this expression called Str_Match 2,621,112 times.
224 1.18 rillig # Adding another '*?' to the pattern called Str_Match 20,630,572 times.
225 1.18 rillig # Adding another '*?' to the pattern called Str_Match 136,405,672 times.
226 1.18 rillig # Adding another '*?' to the pattern called Str_Match 773,168,722 times.
227 1.18 rillig # Adding another '*?' to the pattern called Str_Match 3,815,481,072 times.
228 1.18 rillig # Since 2023-06-22, Str_Match no longer backtracks.
229 1.18 rillig .if ${:U..................................................b:M*?*?*?*?*?a}
230 1.11 rillig .endif
231 1.11 rillig
232 1.18 rillig
233 1.18 rillig # 6. Error handling
234 1.18 rillig
235 1.11 rillig # [ Incomplete empty character list, never matches.
236 1.11 rillig WORDS= a a[
237 1.15 rillig # expect+1: warning: Unfinished character list in pattern 'a[' of modifier ':M'
238 1.11 rillig .if ${WORDS:Ma[} != ""
239 1.11 rillig . error
240 1.11 rillig .endif
241 1.11 rillig
242 1.11 rillig # [^ Incomplete negated empty character list, matches any single
243 1.11 rillig # character.
244 1.11 rillig WORDS= a a[ aX
245 1.15 rillig # expect+1: warning: Unfinished character list in pattern 'a[^' of modifier ':M'
246 1.11 rillig .if ${WORDS:Ma[^} != "a[ aX"
247 1.7 rillig . error
248 1.7 rillig .endif
249 1.7 rillig
250 1.11 rillig # [-x1-3 Incomplete character list, matches those elements that can be
251 1.11 rillig # parsed without lookahead.
252 1.11 rillig WORDS= - + x xx 0 1 2 3 4 [x1-3
253 1.15 rillig # expect+1: warning: Unfinished character list in pattern '[-x1-3' of modifier ':M'
254 1.11 rillig .if ${WORDS:M[-x1-3} != "- x 1 2 3"
255 1.11 rillig . error
256 1.11 rillig .endif
257 1.11 rillig
258 1.14 rillig # *[-x1-3 Incomplete character list after a wildcard, matches those
259 1.14 rillig # words that end with one of the characters from the list.
260 1.14 rillig WORDS= - + x xx 0 1 2 3 4 00 01 10 11 000 001 010 011 100 101 110 111 [x1-3
261 1.15 rillig # expect+1: warning: Unfinished character list in pattern '*[-x1-3' of modifier ':M'
262 1.14 rillig .if ${WORDS:M*[-x1-3} != "- x xx 1 2 3 01 11 001 011 101 111 [x1-3"
263 1.14 rillig . warning ${WORDS:M*[-x1-3}
264 1.14 rillig .endif
265 1.14 rillig
266 1.11 rillig # [^-x1-3
267 1.11 rillig # Incomplete negated character list, matches any character
268 1.11 rillig # except those elements that can be parsed without lookahead.
269 1.11 rillig WORDS= - + x xx 0 1 2 3 4 [x1-3
270 1.15 rillig # expect+1: warning: Unfinished character list in pattern '[^-x1-3' of modifier ':M'
271 1.11 rillig .if ${WORDS:M[^-x1-3} != "+ 0 4"
272 1.11 rillig . error
273 1.11 rillig .endif
274 1.11 rillig
275 1.11 rillig # [\ Incomplete character list containing a single '\'.
276 1.11 rillig #
277 1.11 rillig # A word can only end with a backslash if the preceding
278 1.11 rillig # character is a backslash as well; in all other cases the final
279 1.11 rillig # backslash would escape the following space, making the space
280 1.11 rillig # part of the word. Only the very last word of a string can be
281 1.11 rillig # '\', as there is no following space that could be escaped.
282 1.11 rillig WORDS= \\ \a ${:Ux\\}
283 1.18 rillig PATTERN= ${:U?[\\}
284 1.18 rillig # expect+1: warning: Unfinished character list in pattern '?[\' of modifier ':M'
285 1.18 rillig .if ${WORDS:M${PATTERN}} != "\\\\ x\\"
286 1.11 rillig . error
287 1.11 rillig .endif
288 1.11 rillig
289 1.11 rillig # [x- Incomplete character list containing an incomplete character
290 1.11 rillig # range, matches only the 'x'.
291 1.11 rillig WORDS= [x- x x- y
292 1.15 rillig # expect+1: warning: Unfinished character range in pattern '[x-' of modifier ':M'
293 1.11 rillig .if ${WORDS:M[x-} != "x"
294 1.11 rillig . error
295 1.11 rillig .endif
296 1.11 rillig
297 1.11 rillig # [^x- Incomplete negated character list containing an incomplete
298 1.11 rillig # character range; matches each word that does not have an 'x'
299 1.11 rillig # at the position of the character list.
300 1.11 rillig #
301 1.11 rillig # XXX: Even matches strings that are longer than a single
302 1.11 rillig # character.
303 1.11 rillig WORDS= [x- x x- y yyyyy
304 1.15 rillig # expect+1: warning: Unfinished character range in pattern '[^x-' of modifier ':M'
305 1.11 rillig .if ${WORDS:M[^x-} != "[x- y yyyyy"
306 1.11 rillig . error
307 1.11 rillig .endif
308 1.7 rillig
309 1.18 rillig # [:] matches never since the ':' starts the next modifier
310 1.18 rillig # expect+3: warning: Unfinished character list in pattern '[' of modifier ':M'
311 1.18 rillig # expect+2: Unknown modifier "]"
312 1.18 rillig # expect+1: Malformed conditional (${ ${:U\:} ${:U\:\:} :L:M[:]} != ":")
313 1.18 rillig .if ${ ${:U\:} ${:U\:\:} :L:M[:]} != ":"
314 1.7 rillig . error
315 1.18 rillig .else
316 1.7 rillig . error
317 1.7 rillig .endif
318 1.9 rillig
319 1.9 rillig
320 1.18 rillig # 7. Historical bugs
321 1.12 rillig
322 1.12 rillig # Before var.c 1.1031 from 2022-08-24, the following expressions caused an
323 1.12 rillig # out-of-bounds read beyond the indirect ':M' modifiers.
324 1.12 rillig .if ${:U:${:UM\\}} # The ':M' pattern need not be unescaped, the
325 1.12 rillig . error # resulting pattern is '\', it never matches
326 1.12 rillig .endif # anything.
327 1.12 rillig .if ${:U:${:UM\\\:\\}} # The ':M' pattern must be unescaped, the
328 1.12 rillig . error # resulting pattern is ':\', it never matches
329 1.12 rillig .endif # anything.
330