1 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
3 ;;; Confidential and proprietary information of ITA Software, Inc. ;;;
5 ;;; Copyright (c) 2012 ITA Software, Inc. All rights reserved. ;;;
7 ;;; Original author: Scott McKay ;;;
9 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
11 (in-package "PROTO-IMPL")
14 ;;; .proto file parsing
18 (declaim (inline proto-whitespace-char-p))
19 (defun proto-whitespace-char-p (ch)
20 (locally (declare (optimize (speed 3) (safety 0) (debug 0)))
21 (and ch (member ch '(#\space #\tab #\return #\newline)))))
23 (declaim (inline proto-eol-char-p))
24 (defun proto-eol-char-p (ch)
25 (locally (declare (optimize (speed 3) (safety 0) (debug 0)))
26 (and ch (member ch '(#\return #\newline)))))
28 (declaim (inline proto-token-char-p))
29 (defun proto-token-char-p (ch)
30 (locally (declare (optimize (speed 3) (safety 0) (debug 0)))
31 (and ch (or (alpha-char-p ch)
33 (member ch '(#\_ #\.))))))
36 (defun skip-whitespace (stream)
37 "Skip all the whitespace characters that are coming up in the stream."
38 (loop for ch = (peek-char nil stream nil)
39 until (or (null ch) (not (proto-whitespace-char-p ch)))
40 do (read-char stream nil)))
42 ;;--- Collect the comment so we can attach it to its associated object
43 (defun maybe-skip-comments (stream)
44 "If what appears next in the stream is a comment, skip it and any following comments,
45 then skip any following whitespace."
47 (unless (eql (peek-char nil stream nil) #\/)
50 (case (peek-char nil stream nil)
52 (skip-line-comment stream))
54 (skip-block-comment stream))
56 (error "Found a '~C' at position ~D to start a comment, but no following '~C' or '~C'"
57 #\/ (file-position stream) #\/ #\*)))))
58 (skip-whitespace stream))
60 (defun skip-line-comment (stream)
61 "Skip to the end of a line comment, that is, to the end of the line.
62 Then skip any following whitespace."
63 (loop for ch = (read-char stream nil)
64 until (or (null ch) (proto-eol-char-p ch)))
65 (skip-whitespace stream))
67 (defun skip-block-comment (stream)
68 "Skip to the end of a block comment, that is, until a '*/' is seen.
69 Then skip any following whitespace."
70 (loop for ch = (read-char stream nil)
72 (error "Premature end of file while skipping block comment"))
74 (eql (peek-char nil stream nil) #\/))
75 (read-char stream nil)
77 (skip-whitespace stream))
80 (defun expect-char (stream ch &optional within)
81 "Expect to see 'ch' as the next character in the stream; signal an error if it's not there.
82 Then skip all of the following whitespace."
84 (member (peek-char nil stream) ch)
85 (eql (peek-char nil stream) ch))
87 (error "No '~C' found~@[ within '~A'~] at position ~D"
88 ch within (file-position stream)))
89 (skip-whitespace stream))
92 (defun parse-token (stream)
93 "Parse the next token in the stream, then skip the following whitespace.
94 The returned value is the token."
95 (when (proto-token-char-p (peek-char nil stream nil))
96 (loop for ch = (read-char stream nil)
97 for ch1 = (peek-char nil stream nil)
99 until (or (null ch1) (not (proto-token-char-p ch1)))
101 (skip-whitespace stream)
102 (return (coerce token 'string))))))
104 (defun parse-string (stream)
105 "Parse the next quoted string in the stream, then skip the following whitespace.
106 The returned value is the string, without the quotation marks."
107 (loop with ch0 = (read-char stream nil)
108 for ch = (read-char stream nil)
109 until (or (null ch) (char= ch ch0))
110 collect ch into string
112 (skip-whitespace stream)
113 (return (coerce string 'string)))))
115 (defun parse-int (stream)
116 "Parse the next token in the stream as an integer, then skip the following whitespace.
117 The returned value is the integer."
118 (when (digit-char-p (peek-char nil stream nil))
119 (loop for ch = (read-char stream nil)
120 for ch1 = (peek-char nil stream nil)
121 collect ch into token
122 until (or (null ch1) (not (digit-char-p ch1)))
124 (skip-whitespace stream)
125 (return (parse-integer (coerce token 'string)))))))
127 (defun parse-float (stream)
128 "Parse the next token in the stream as a float, then skip the following whitespace. The returned value is the float."
129 (when (let ((ch (peek-char nil stream nil)))
130 (or (digit-char-p ch) (eql ch #\-)))
131 (let ((token (parse-token stream)))
133 (skip-whitespace stream)
134 (coerce (read-from-string token) 'float)))))
137 ;;; The parser itself
139 (defun parse-protobuf-from-file (filename)
140 "Parses the named file as a .proto file, and returns the Protobufs schema."
141 (with-open-file (stream filename
143 :external-format :utf-8
144 :element-type 'character)
145 (parse-protobuf-from-stream stream
146 :name (class-name->proto (pathname-name (pathname stream)))
147 :class (pathname-name (pathname stream)))))
149 ;; The syntax for Protocol Buffers is so simple that it doesn't seem worth
150 ;; writing a sophisticated parser
151 ;; Note that we don't put the result into *all-protobufs*; do that at a higher level
152 (defun parse-protobuf-from-stream (stream &key name class)
153 "Parses a top-level .proto file from the stream 'stream'.
154 Returns the protobuf schema that describes the .proto file."
155 (let* ((protobuf (make-instance 'protobuf
158 (*protobuf* protobuf)
159 (*protobuf-package* nil))
161 (skip-whitespace stream)
162 (maybe-skip-comments stream)
163 (let ((char (peek-char nil stream nil)))
165 (return-from parse-protobuf-from-stream protobuf))
166 ((proto-token-char-p char)
167 (let ((token (parse-token stream)))
168 (cond ((string= token "syntax")
169 (parse-proto-syntax stream protobuf))
170 ((string= token "package")
171 (parse-proto-package stream protobuf))
172 ((string= token "import")
173 (parse-proto-import stream protobuf))
174 ((string= token "option")
175 (parse-proto-option stream protobuf))
176 ((string= token "enum")
177 (parse-proto-enum stream protobuf))
178 ((string= token "message")
179 (parse-proto-message stream protobuf))
180 ((string= token "service")
181 (parse-proto-service stream protobuf)))))
183 (error "Syntax error at position ~D" (file-position stream))))))))
185 (defun parse-proto-syntax (stream protobuf &optional (terminator #\;))
186 "Parse a Protobufs syntax line from 'stream'.
187 Updates the 'protobuf' object to use the syntax."
188 (let ((syntax (prog1 (parse-token stream)
189 (expect-char stream terminator "syntax")
190 (maybe-skip-comments stream))))
191 (setf (proto-syntax protobuf) syntax)))
193 (defun parse-proto-package (stream protobuf &optional (terminator #\;))
194 "Parse a Protobufs package line from 'stream'.
195 Updates the 'protobuf' object to use the package."
196 (let* ((package (prog1 (substitute #\- #\_ (parse-token stream))
197 (expect-char stream terminator "package")
198 (maybe-skip-comments stream)))
199 (lisp-pkg (or (find-package package)
200 (find-package (string-upcase package)))))
201 (setq *protobuf-package* lisp-pkg)
202 (setf (proto-package protobuf) package)))
204 (defun parse-proto-import (stream protobuf &optional (terminator #\;))
205 "Parse a Protobufs import line from 'stream'.
206 Updates the 'protobuf' object to use the package."
207 (let ((import (prog1 (parse-string stream)
208 (expect-char stream terminator "package")
209 (maybe-skip-comments stream))))
210 (setf (proto-imports protobuf) (nconc (proto-imports protobuf) (list import)))))
212 (defun parse-proto-option (stream protobuf &optional (terminator #\;))
213 "Parse a Protobufs option from 'stream'.
214 Updates the 'protobuf' (or message, service, RPC) to have the option."
215 (let* ((key (prog1 (parse-token stream)
216 (expect-char stream #\= "option")))
217 (val (prog1 (if (eql (peek-char nil stream nil) #\")
218 (parse-string stream)
219 (parse-token stream))
220 (expect-char stream terminator "option")
221 (maybe-skip-comments stream)))
222 (option (make-instance 'protobuf-option
226 (setf (proto-options protobuf) (nconc (proto-options protobuf) (list option)))
227 (when (and (string= key "optimize_for")
228 (typep protobuf 'protobuf))
229 (let ((value (cond ((string= val "SPEED") :speed)
230 ((string= val "CODE_SIZE") :space)
232 (setf (proto-optimize protobuf) value))))
234 ;; If nothing to graft the option into, just return it as the value
238 (defun parse-proto-enum (stream protobuf)
239 "Parse a Protobufs enum from 'stream'.
240 Updates the 'protobuf' or 'protobuf-message' object to have the enum."
241 (let* ((name (prog1 (parse-token stream)
242 (expect-char stream #\{ "enum")
243 (maybe-skip-comments stream)))
244 (enum (make-instance 'protobuf-enum
245 :class (proto->class-name name *protobuf-package*)
248 (let ((name (parse-token stream)))
250 (expect-char stream #\} "enum")
251 (maybe-skip-comments stream)
252 (setf (proto-enums protobuf) (nconc (proto-messages protobuf) (list enum)))
253 (let ((type (find-option enum "lisp_name")))
255 (setf (proto-class enum) (make-lisp-symbol type))))
256 (let ((alias (find-option enum "lisp_alias")))
258 (setf (proto-alias-for enum) (make-lisp-symbol alias))))
259 (return-from parse-proto-enum))
260 (if (string= name "option")
261 (parse-proto-option stream enum #\;)
262 (parse-proto-enum-value stream enum name))))))
264 (defun parse-proto-enum-value (stream enum name)
265 "Parse a Protobufs enum vvalue from 'stream'.
266 Updates the 'protobuf-enum' object to have the enum value."
267 (expect-char stream #\= "enum")
268 (let* ((idx (prog1 (parse-int stream)
269 (expect-char stream #\; "enum")
270 (maybe-skip-comments stream)))
271 (value (make-instance 'protobuf-enum-value
274 :value (proto->enum-name name *protobuf-package*))))
275 (setf (proto-values enum) (nconc (proto-values enum) (list value)))))
278 (defun parse-proto-message (stream protobuf)
279 "Parse a Protobufs message from 'stream'.
280 Updates the 'protobuf' or 'protobuf-message' object to have the message."
281 (let* ((name (prog1 (parse-token stream)
282 (expect-char stream #\{ "message")
283 (maybe-skip-comments stream)))
284 (message (make-instance 'protobuf-message
285 :class (proto->class-name name *protobuf-package*)
288 (let ((token (parse-token stream)))
290 (expect-char stream #\} "message")
291 (maybe-skip-comments stream)
292 (setf (proto-messages protobuf) (nconc (proto-messages protobuf) (list message)))
293 (let ((type (find-option message "lisp_name")))
295 (setf (proto-class message) (make-lisp-symbol type))))
296 (let ((alias (find-option message "lisp_alias")))
298 (setf (proto-alias-for message) (make-lisp-symbol alias))))
299 (return-from parse-proto-message))
300 (cond ((string= token "enum")
301 (parse-proto-enum stream message))
302 ((string= token "message")
303 (parse-proto-message stream message))
304 ((string= token "option")
305 (parse-proto-option stream message #\;))
306 ((member token '("required" "optional" "repeated") :test #'string=)
307 (parse-proto-field stream message token))
309 (error "Unrecognized token ~A at position ~D"
310 token (file-position stream))))))))
312 (defun parse-proto-field (stream message required)
313 "Parse a Protobufs field from 'stream'.
314 Updates the 'protobuf-message' object to have the field."
315 (let* ((type (parse-token stream))
316 (name (prog1 (parse-token stream)
317 (expect-char stream #\= "message")))
318 (idx (parse-int stream))
319 (opts (prog1 (parse-proto-field-options stream)
320 (expect-char stream #\; "message")
321 (maybe-skip-comments stream)))
322 (dflt (find-option opts "default"))
323 (packed (find-option opts "packed"))
324 (ptype (if (member type '("int32" "int64" "uint32" "uint64" "sint32" "sint64"
325 "fixed32" "fixed64" "sfixed32" "sfixed64"
326 "string" "bytes" "bool" "float" "double") :test #'string=)
329 (class (if (keywordp ptype) ptype (proto->class-name type *protobuf-package*)))
330 (field (make-instance 'protobuf-field
332 :value (proto->slot-name name *protobuf-package*)
335 ;; One of :required, :optional or :repeated
336 :required (kintern required)
339 :packed (and packed (string= packed "true")))))
340 (let ((slot (find-option opts "lisp_name")))
342 (setf (proto-value field) (make-lisp-symbol type))))
343 (setf (proto-fields message) (nconc (proto-fields message) (list field)))))
345 (defun parse-proto-field-options (stream)
346 "Parse any options in a Protobufs field from 'stream'.
347 Returns a list of 'protobuf-option' objects."
348 (with-collectors ((options collect-option))
350 (unless (eql (peek-char nil stream) #\[)
351 (return-from parse-proto-field-options options))
352 (expect-char stream #\[ "message")
353 (collect-option (parse-proto-option stream nil #\])))
357 (defun parse-proto-service (stream protobuf)
358 "Parse a Protobufs service from 'stream'.
359 Updates the 'protobuf-message' object to have the service."
360 (let* ((name (prog1 (parse-token stream)
361 (expect-char stream #\{ "service")
362 (maybe-skip-comments stream)))
363 (service (make-instance 'protobuf-service
364 :class (proto->class-name name *protobuf-package*)
367 (let ((token (parse-token stream)))
369 (expect-char stream #\} "service")
370 (maybe-skip-comments stream)
371 (setf (proto-services protobuf) (nconc (proto-services protobuf) (list service)))
372 (return-from parse-proto-service))
373 (cond ((string= token "option")
374 (parse-proto-option stream service #\;))
375 ((string= token "rpc")
376 (parse-proto-rpc stream service token))
378 (error "Unrecognized token ~A at position ~D"
379 token (file-position stream))))))))
381 (defun parse-proto-rpc (stream service rpc)
382 "Parse a Protobufs enum vvalue from 'stream'.
383 Updates the 'protobuf-enum' object to have the enum value."
384 (declare (ignore rpc))
385 (let* ((name (parse-token stream))
386 (in (prog2 (expect-char stream #\( "service")
388 (expect-char stream #\) "service")))
389 (ret (parse-token stream))
390 (out (prog2 (expect-char stream #\( "service")
392 (expect-char stream #\) "service")))
393 (opts (let ((opts (parse-proto-rpc-options stream)))
394 (when (or (null opts) (eql (peek-char nil stream) #\;))
395 (expect-char stream #\; "service"))
396 (maybe-skip-comments stream)
398 (rpc (make-instance 'protobuf-rpc
399 :class (proto->class-name name *protobuf-package*)
401 :input-type (proto->class-name in *protobuf-package*)
403 :output-type (proto->class-name out *protobuf-package*)
406 (let ((name (find-option rpc "lisp_name")))
408 (setf (proto-function rpc) (make-lisp-symbol name))))
409 (assert (string= ret "returns") ()
410 "Syntax error in 'message' at position ~D" (file-position stream))
411 (setf (proto-rpcs service) (nconc (proto-rpcs service) (list rpc)))))
413 (defun parse-proto-rpc-options (stream)
414 "Parse any options in a Protobufs RPC from 'stream'.
415 Returns a list of 'protobuf-option' objects."
416 (when (eql (peek-char nil stream) #\{)
417 (expect-char stream #\{ "service")
418 (maybe-skip-comments stream)
419 (with-collectors ((options collect-option))
421 (when (eql (peek-char nil stream) #\})
423 (assert (string= (parse-token stream) "option") ()
424 "Syntax error in 'message' at position ~D" (file-position stream))
425 (collect-option (parse-proto-option stream nil #\;)))
426 (expect-char stream #\} "service")
427 (maybe-skip-comments stream)