• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // Protocol Buffers - Google's data interchange format
2 // Copyright 2008 Google Inc.  All rights reserved.
3 // https://developers.google.com/protocol-buffers/
4 //
5 // Redistribution and use in source and binary forms, with or without
6 // modification, are permitted provided that the following conditions are
7 // met:
8 //
9 //     * Redistributions of source code must retain the above copyright
10 // notice, this list of conditions and the following disclaimer.
11 //     * Redistributions in binary form must reproduce the above
12 // copyright notice, this list of conditions and the following disclaimer
13 // in the documentation and/or other materials provided with the
14 // distribution.
15 //     * Neither the name of Google Inc. nor the names of its
16 // contributors may be used to endorse or promote products derived from
17 // this software without specific prior written permission.
18 //
19 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 
31 // Author: kenton@google.com (Kenton Varda)
32 //  Based on original Protocol Buffers design by
33 //  Sanjay Ghemawat, Jeff Dean, and others.
34 //
35 // Implements the Protocol Compiler front-end such that it may be reused by
36 // custom compilers written to support other languages.
37 
38 #ifndef GOOGLE_PROTOBUF_COMPILER_COMMAND_LINE_INTERFACE_H__
39 #define GOOGLE_PROTOBUF_COMPILER_COMMAND_LINE_INTERFACE_H__
40 
41 #include <map>
42 #include <set>
43 #include <string>
44 #include <unordered_map>
45 #include <utility>
46 #include <vector>
47 
48 #include <google/protobuf/stubs/common.h>
49 
50 #include <google/protobuf/port_def.inc>
51 
52 namespace google {
53 namespace protobuf {
54 
55 class Descriptor;           // descriptor.h
56 class DescriptorDatabase;   // descriptor_database.h
57 class DescriptorPool;       // descriptor.h
58 class FileDescriptor;       // descriptor.h
59 class FileDescriptorSet;    // descriptor.h
60 class FileDescriptorProto;  // descriptor.pb.h
61 template <typename T>
62 class RepeatedPtrField;          // repeated_field.h
63 class SimpleDescriptorDatabase;  // descriptor_database.h
64 
65 namespace compiler {
66 
67 class CodeGenerator;     // code_generator.h
68 class GeneratorContext;  // code_generator.h
69 class DiskSourceTree;    // importer.h
70 
71 // This class implements the command-line interface to the protocol compiler.
72 // It is designed to make it very easy to create a custom protocol compiler
73 // supporting the languages of your choice.  For example, if you wanted to
74 // create a custom protocol compiler binary which includes both the regular
75 // C++ support plus support for your own custom output "Foo", you would
76 // write a class "FooGenerator" which implements the CodeGenerator interface,
77 // then write a main() procedure like this:
78 //
79 //   int main(int argc, char* argv[]) {
80 //     google::protobuf::compiler::CommandLineInterface cli;
81 //
82 //     // Support generation of C++ source and headers.
83 //     google::protobuf::compiler::cpp::CppGenerator cpp_generator;
84 //     cli.RegisterGenerator("--cpp_out", &cpp_generator,
85 //       "Generate C++ source and header.");
86 //
87 //     // Support generation of Foo code.
88 //     FooGenerator foo_generator;
89 //     cli.RegisterGenerator("--foo_out", &foo_generator,
90 //       "Generate Foo file.");
91 //
92 //     return cli.Run(argc, argv);
93 //   }
94 //
95 // The compiler is invoked with syntax like:
96 //   protoc --cpp_out=outdir --foo_out=outdir --proto_path=src src/foo.proto
97 //
98 // The .proto file to compile can be specified on the command line using either
99 // its physical file path, or a virtual path relative to a diretory specified
100 // in --proto_path. For example, for src/foo.proto, the following two protoc
101 // invocations work the same way:
102 //   1. protoc --proto_path=src src/foo.proto (physical file path)
103 //   2. protoc --proto_path=src foo.proto (virtual path relative to src)
104 //
105 // If a file path can be interpreted both as a physical file path and as a
106 // relative virtual path, the physical file path takes precendence.
107 //
108 // For a full description of the command-line syntax, invoke it with --help.
109 class PROTOC_EXPORT CommandLineInterface {
110  public:
111   static const char* const kPathSeparator;
112 
113   CommandLineInterface();
114   ~CommandLineInterface();
115 
116   // Register a code generator for a language.
117   //
118   // Parameters:
119   // * flag_name: The command-line flag used to specify an output file of
120   //   this type.  The name must start with a '-'.  If the name is longer
121   //   than one letter, it must start with two '-'s.
122   // * generator: The CodeGenerator which will be called to generate files
123   //   of this type.
124   // * help_text: Text describing this flag in the --help output.
125   //
126   // Some generators accept extra parameters.  You can specify this parameter
127   // on the command-line by placing it before the output directory, separated
128   // by a colon:
129   //   protoc --foo_out=enable_bar:outdir
130   // The text before the colon is passed to CodeGenerator::Generate() as the
131   // "parameter".
132   void RegisterGenerator(const std::string& flag_name, CodeGenerator* generator,
133                          const std::string& help_text);
134 
135   // Register a code generator for a language.
136   // Besides flag_name you can specify another option_flag_name that could be
137   // used to pass extra parameters to the registered code generator.
138   // Suppose you have registered a generator by calling:
139   //   command_line_interface.RegisterGenerator("--foo_out", "--foo_opt", ...)
140   // Then you could invoke the compiler with a command like:
141   //   protoc --foo_out=enable_bar:outdir --foo_opt=enable_baz
142   // This will pass "enable_bar,enable_baz" as the parameter to the generator.
143   void RegisterGenerator(const std::string& flag_name,
144                          const std::string& option_flag_name,
145                          CodeGenerator* generator,
146                          const std::string& help_text);
147 
148   // Enables "plugins".  In this mode, if a command-line flag ends with "_out"
149   // but does not match any registered generator, the compiler will attempt to
150   // find a "plugin" to implement the generator.  Plugins are just executables.
151   // They should live somewhere in the PATH.
152   //
153   // The compiler determines the executable name to search for by concatenating
154   // exe_name_prefix with the unrecognized flag name, removing "_out".  So, for
155   // example, if exe_name_prefix is "protoc-" and you pass the flag --foo_out,
156   // the compiler will try to run the program "protoc-foo".
157   //
158   // The plugin program should implement the following usage:
159   //   plugin [--out=OUTDIR] [--parameter=PARAMETER] PROTO_FILES < DESCRIPTORS
160   // --out indicates the output directory (as passed to the --foo_out
161   // parameter); if omitted, the current directory should be used.  --parameter
162   // gives the generator parameter, if any was provided (see below).  The
163   // PROTO_FILES list the .proto files which were given on the compiler
164   // command-line; these are the files for which the plugin is expected to
165   // generate output code.  Finally, DESCRIPTORS is an encoded FileDescriptorSet
166   // (as defined in descriptor.proto).  This is piped to the plugin's stdin.
167   // The set will include descriptors for all the files listed in PROTO_FILES as
168   // well as all files that they import.  The plugin MUST NOT attempt to read
169   // the PROTO_FILES directly -- it must use the FileDescriptorSet.
170   //
171   // The plugin should generate whatever files are necessary, as code generators
172   // normally do.  It should write the names of all files it generates to
173   // stdout.  The names should be relative to the output directory, NOT absolute
174   // names or relative to the current directory.  If any errors occur, error
175   // messages should be written to stderr.  If an error is fatal, the plugin
176   // should exit with a non-zero exit code.
177   //
178   // Plugins can have generator parameters similar to normal built-in
179   // generators. Extra generator parameters can be passed in via a matching
180   // "_opt" parameter. For example:
181   //   protoc --plug_out=enable_bar:outdir --plug_opt=enable_baz
182   // This will pass "enable_bar,enable_baz" as the parameter to the plugin.
183   //
184   void AllowPlugins(const std::string& exe_name_prefix);
185 
186   // Run the Protocol Compiler with the given command-line parameters.
187   // Returns the error code which should be returned by main().
188   //
189   // It may not be safe to call Run() in a multi-threaded environment because
190   // it calls strerror().  I'm not sure why you'd want to do this anyway.
191   int Run(int argc, const char* const argv[]);
192 
193   // DEPRECATED. Calling this method has no effect. Protocol compiler now
194   // always try to find the .proto file relative to the current directory
195   // first and if the file is not found, it will then treat the input path
196   // as a virutal path.
SetInputsAreProtoPathRelative(bool)197   void SetInputsAreProtoPathRelative(bool /* enable */) {}
198 
199   // Provides some text which will be printed when the --version flag is
200   // used.  The version of libprotoc will also be printed on the next line
201   // after this text.
SetVersionInfo(const std::string & text)202   void SetVersionInfo(const std::string& text) { version_info_ = text; }
203 
204 
205  private:
206   // -----------------------------------------------------------------
207 
208   class ErrorPrinter;
209   class GeneratorContextImpl;
210   class MemoryOutputStream;
211   typedef std::unordered_map<std::string, GeneratorContextImpl*>
212       GeneratorContextMap;
213 
214   // Clear state from previous Run().
215   void Clear();
216 
217   // Remaps the proto file so that it is relative to one of the directories
218   // in proto_path_.  Returns false if an error occurred.
219   bool MakeProtoProtoPathRelative(DiskSourceTree* source_tree,
220                                   std::string* proto,
221                                   DescriptorDatabase* fallback_database);
222 
223   // Remaps each file in input_files_ so that it is relative to one of the
224   // directories in proto_path_.  Returns false if an error occurred.
225   bool MakeInputsBeProtoPathRelative(DiskSourceTree* source_tree,
226                                      DescriptorDatabase* fallback_database);
227 
228 
229   // Return status for ParseArguments() and InterpretArgument().
230   enum ParseArgumentStatus {
231     PARSE_ARGUMENT_DONE_AND_CONTINUE,
232     PARSE_ARGUMENT_DONE_AND_EXIT,
233     PARSE_ARGUMENT_FAIL
234   };
235 
236   // Parse all command-line arguments.
237   ParseArgumentStatus ParseArguments(int argc, const char* const argv[]);
238 
239   // Read an argument file and append the file's content to the list of
240   // arguments. Return false if the file cannot be read.
241   bool ExpandArgumentFile(const std::string& file,
242                           std::vector<std::string>* arguments);
243 
244   // Parses a command-line argument into a name/value pair.  Returns
245   // true if the next argument in the argv should be used as the value,
246   // false otherwise.
247   //
248   // Examples:
249   //   "-Isrc/protos" ->
250   //     name = "-I", value = "src/protos"
251   //   "--cpp_out=src/foo.pb2.cc" ->
252   //     name = "--cpp_out", value = "src/foo.pb2.cc"
253   //   "foo.proto" ->
254   //     name = "", value = "foo.proto"
255   bool ParseArgument(const char* arg, std::string* name, std::string* value);
256 
257   // Interprets arguments parsed with ParseArgument.
258   ParseArgumentStatus InterpretArgument(const std::string& name,
259                                         const std::string& value);
260 
261   // Print the --help text to stderr.
262   void PrintHelpText();
263 
264   // Loads proto_path_ into the provided source_tree.
265   bool InitializeDiskSourceTree(DiskSourceTree* source_tree,
266                                 DescriptorDatabase* fallback_database);
267 
268   // Verify that all the input files exist in the given database.
269   bool VerifyInputFilesInDescriptors(DescriptorDatabase* fallback_database);
270 
271   // Loads descriptor_set_in into the provided database
272   bool PopulateSimpleDescriptorDatabase(SimpleDescriptorDatabase* database);
273 
274   // Parses input_files_ into parsed_files
275   bool ParseInputFiles(DescriptorPool* descriptor_pool,
276                        std::vector<const FileDescriptor*>* parsed_files);
277 
278   // Generate the given output file from the given input.
279   struct OutputDirective;  // see below
280   bool GenerateOutput(const std::vector<const FileDescriptor*>& parsed_files,
281                       const OutputDirective& output_directive,
282                       GeneratorContext* generator_context);
283   bool GeneratePluginOutput(
284       const std::vector<const FileDescriptor*>& parsed_files,
285       const std::string& plugin_name, const std::string& parameter,
286       GeneratorContext* generator_context, std::string* error);
287 
288   // Implements --encode and --decode.
289   bool EncodeOrDecode(const DescriptorPool* pool);
290 
291   // Implements the --descriptor_set_out option.
292   bool WriteDescriptorSet(
293       const std::vector<const FileDescriptor*>& parsed_files);
294 
295   // Implements the --dependency_out option
296   bool GenerateDependencyManifestFile(
297       const std::vector<const FileDescriptor*>& parsed_files,
298       const GeneratorContextMap& output_directories,
299       DiskSourceTree* source_tree);
300 
301   // Get all transitive dependencies of the given file (including the file
302   // itself), adding them to the given list of FileDescriptorProtos.  The
303   // protos will be ordered such that every file is listed before any file that
304   // depends on it, so that you can call DescriptorPool::BuildFile() on them
305   // in order.  Any files in *already_seen will not be added, and each file
306   // added will be inserted into *already_seen.  If include_source_code_info is
307   // true then include the source code information in the FileDescriptorProtos.
308   // If include_json_name is true, populate the json_name field of
309   // FieldDescriptorProto for all fields.
310   static void GetTransitiveDependencies(
311       const FileDescriptor* file, bool include_json_name,
312       bool include_source_code_info,
313       std::set<const FileDescriptor*>* already_seen,
314       RepeatedPtrField<FileDescriptorProto>* output);
315 
316   // Implements the --print_free_field_numbers. This function prints free field
317   // numbers into stdout for the message and it's nested message types in
318   // post-order, i.e. nested types first. Printed range are left-right
319   // inclusive, i.e. [a, b].
320   //
321   // Groups:
322   // For historical reasons, groups are considered to share the same
323   // field number space with the parent message, thus it will not print free
324   // field numbers for groups. The field numbers used in the groups are
325   // excluded in the free field numbers of the parent message.
326   //
327   // Extension Ranges:
328   // Extension ranges are considered ocuppied field numbers and they will not be
329   // listed as free numbers in the output.
330   void PrintFreeFieldNumbers(const Descriptor* descriptor);
331 
332   // -----------------------------------------------------------------
333 
334   // The name of the executable as invoked (i.e. argv[0]).
335   std::string executable_name_;
336 
337   // Version info set with SetVersionInfo().
338   std::string version_info_;
339 
340   // Registered generators.
341   struct GeneratorInfo {
342     std::string flag_name;
343     std::string option_flag_name;
344     CodeGenerator* generator;
345     std::string help_text;
346   };
347   typedef std::map<std::string, GeneratorInfo> GeneratorMap;
348   GeneratorMap generators_by_flag_name_;
349   GeneratorMap generators_by_option_name_;
350   // A map from generator names to the parameters specified using the option
351   // flag. For example, if the user invokes the compiler with:
352   //   protoc --foo_out=outputdir --foo_opt=enable_bar ...
353   // Then there will be an entry ("--foo_out", "enable_bar") in this map.
354   std::map<std::string, std::string> generator_parameters_;
355   // Similar to generator_parameters_, but stores the parameters for plugins.
356   std::map<std::string, std::string> plugin_parameters_;
357 
358   // See AllowPlugins().  If this is empty, plugins aren't allowed.
359   std::string plugin_prefix_;
360 
361   // Maps specific plugin names to files.  When executing a plugin, this map
362   // is searched first to find the plugin executable.  If not found here, the
363   // PATH (or other OS-specific search strategy) is searched.
364   std::map<std::string, std::string> plugins_;
365 
366   // Stuff parsed from command line.
367   enum Mode {
368     MODE_COMPILE,  // Normal mode:  parse .proto files and compile them.
369     MODE_ENCODE,   // --encode:  read text from stdin, write binary to stdout.
370     MODE_DECODE,   // --decode:  read binary from stdin, write text to stdout.
371     MODE_PRINT,    // Print mode: print info of the given .proto files and exit.
372   };
373 
374   Mode mode_;
375 
376   enum PrintMode {
377     PRINT_NONE,         // Not in MODE_PRINT
378     PRINT_FREE_FIELDS,  // --print_free_fields
379   };
380 
381   PrintMode print_mode_;
382 
383   enum ErrorFormat {
384     ERROR_FORMAT_GCC,  // GCC error output format (default).
385     ERROR_FORMAT_MSVS  // Visual Studio output (--error_format=msvs).
386   };
387 
388   ErrorFormat error_format_;
389 
390   std::vector<std::pair<std::string, std::string> >
391       proto_path_;                        // Search path for proto files.
392   std::vector<std::string> input_files_;  // Names of the input proto files.
393 
394   // Names of proto files which are allowed to be imported. Used by build
395   // systems to enforce depend-on-what-you-import.
396   std::set<std::string> direct_dependencies_;
397   bool direct_dependencies_explicitly_set_;
398 
399   // If there's a violation of depend-on-what-you-import, this string will be
400   // presented to the user. "%s" will be replaced with the violating import.
401   std::string direct_dependencies_violation_msg_;
402 
403   // output_directives_ lists all the files we are supposed to output and what
404   // generator to use for each.
405   struct OutputDirective {
406     std::string name;          // E.g. "--foo_out"
407     CodeGenerator* generator;  // NULL for plugins
408     std::string parameter;
409     std::string output_location;
410   };
411   std::vector<OutputDirective> output_directives_;
412 
413   // When using --encode or --decode, this names the type we are encoding or
414   // decoding.  (Empty string indicates --decode_raw.)
415   std::string codec_type_;
416 
417   // If --descriptor_set_in was given, these are filenames containing
418   // parsed FileDescriptorSets to be used for loading protos.  Otherwise, empty.
419   std::vector<std::string> descriptor_set_in_names_;
420 
421   // If --descriptor_set_out was given, this is the filename to which the
422   // FileDescriptorSet should be written.  Otherwise, empty.
423   std::string descriptor_set_out_name_;
424 
425   // If --dependency_out was given, this is the path to the file where the
426   // dependency file will be written. Otherwise, empty.
427   std::string dependency_out_name_;
428 
429   // True if --include_imports was given, meaning that we should
430   // write all transitive dependencies to the DescriptorSet.  Otherwise, only
431   // the .proto files listed on the command-line are added.
432   bool imports_in_descriptor_set_;
433 
434   // True if --include_source_info was given, meaning that we should not strip
435   // SourceCodeInfo from the DescriptorSet.
436   bool source_info_in_descriptor_set_;
437 
438   // Was the --disallow_services flag used?
439   bool disallow_services_;
440 
441   GOOGLE_DISALLOW_EVIL_CONSTRUCTORS(CommandLineInterface);
442 };
443 
444 }  // namespace compiler
445 }  // namespace protobuf
446 }  // namespace google
447 
448 #include <google/protobuf/port_undef.inc>
449 
450 #endif  // GOOGLE_PROTOBUF_COMPILER_COMMAND_LINE_INTERFACE_H__
451