tokenize.h

// Copyright 2020 The Pigweed Authors
//
// Licensed under the Apache License, Version 2.0 (the "License"); you may not
// use this file except in compliance with the License. You may obtain a copy of
// the License at
//
//     https://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
// WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
// License for the specific language governing permissions and limitations under
// the License.
#pragma once
 
#ifdef __cplusplus
 
#include <cstddef>
#include <cstdint>
#include <string_view>
 
#else
 
#include <stddef.h>
#include <stdint.h>
 
#endif  // __cplusplus
 
#include "pw_polyfill/static_assert.h"
#include "pw_preprocessor/arguments.h"
#include "pw_preprocessor/compiler.h"
#include "pw_preprocessor/concat.h"
#include "pw_preprocessor/util.h"
#include "pw_tokenizer/internal/argument_types.h"
#include "pw_tokenizer/internal/tokenize_string.h"
 
typedef uint32_t pw_tokenizer_Token;
 
// Strings may optionally be tokenized to a domain. Strings in different
// domains can be processed separately by the token database tools. Each domain
// in use must have a corresponding section declared in the linker script. See
// `pw_tokenizer_linker_sections.ld` for more details.
//
// The default domain is an empty string.
#define PW_TOKENIZER_DEFAULT_DOMAIN ""
 
#define PW_TOKENIZE_STRING(...) \
  PW_DELEGATE_BY_ARG_COUNT(_PW_TOKENIZE_STRING_, __VA_ARGS__)
 
#define _PW_TOKENIZE_STRING_1(string_literal) \
  PW_TOKENIZE_STRING_DOMAIN(PW_TOKENIZER_DEFAULT_DOMAIN, string_literal)
 
#define _PW_TOKENIZE_STRING_2(domain, string_literal) \
  PW_TOKENIZE_STRING_DOMAIN(domain, string_literal)
 
#define PW_TOKENIZE_STRING_EXPR(...) \
  PW_DELEGATE_BY_ARG_COUNT(_PW_TOKENIZE_STRING_EXPR_, __VA_ARGS__)
 
#define _PW_TOKENIZE_STRING_EXPR_1(string_literal) \
  _PW_TOKENIZE_STRING_EXPR_2(PW_TOKENIZER_DEFAULT_DOMAIN, string_literal)
 
#define _PW_TOKENIZE_STRING_EXPR_2(domain, string_literal) \
  [&] {                                                    \
    constexpr uint32_t lambda_ret_token =                  \
        PW_TOKENIZE_STRING_DOMAIN(domain, string_literal); \
    return lambda_ret_token;                               \
  }()
 
#define PW_TOKENIZE_STRING_DOMAIN(domain, string_literal) \
  PW_TOKENIZE_STRING_MASK(domain, UINT32_MAX, string_literal)
 
#define PW_TOKENIZE_STRING_DOMAIN_EXPR(domain, string_literal) \
  [&] {                                                        \
    constexpr uint32_t lambda_ret_token =                      \
        PW_TOKENIZE_STRING_DOMAIN(domain, string_literal);     \
    return lambda_ret_token;                                   \
  }()
 
#define PW_TOKENIZE_STRING_MASK(domain, mask, string_literal)                \
  /* assign to a variable */ _PW_TOKENIZER_MASK_TOKEN(mask, string_literal); \
                                                                             \
  static_assert(0 < (mask) && (mask) <= UINT32_MAX,                          \
                "Tokenizer masks must be non-zero uint32_t values.");        \
                                                                             \
  PW_TOKENIZER_DEFINE_TOKEN(                                                 \
      _PW_TOKENIZER_MASK_TOKEN(mask, string_literal), domain, string_literal)
 
#define PW_TOKENIZE_STRING_MASK_EXPR(domain, mask, string_literal) \
  [&] {                                                            \
    constexpr uint32_t lambda_ret_token =                          \
        PW_TOKENIZE_STRING_MASK(domain, mask, string_literal);     \
    return lambda_ret_token;                                       \
  }()
 
#define _PW_TOKENIZER_MASK_TOKEN(mask, string_literal) \
  ((pw_tokenizer_Token)(mask) & PW_TOKENIZER_STRING_TOKEN(string_literal))
 
#define PW_TOKENIZE_TO_BUFFER(buffer, buffer_size_pointer, format, ...) \
  PW_TOKENIZE_TO_BUFFER_DOMAIN(PW_TOKENIZER_DEFAULT_DOMAIN,             \
                               buffer,                                  \
                               buffer_size_pointer,                     \
                               format,                                  \
                               __VA_ARGS__)
 
#define PW_TOKENIZE_TO_BUFFER_DOMAIN(                 \
    domain, buffer, buffer_size_pointer, format, ...) \
  PW_TOKENIZE_TO_BUFFER_MASK(                         \
      domain, UINT32_MAX, buffer, buffer_size_pointer, format, __VA_ARGS__)
 
#define PW_TOKENIZE_TO_BUFFER_MASK(                                          \
    domain, mask, buffer, buffer_size_pointer, format, ...)                  \
  do {                                                                       \
    PW_TOKENIZE_FORMAT_STRING(domain, mask, format, __VA_ARGS__);            \
    _pw_tokenizer_ToBuffer(buffer,                                           \
                           buffer_size_pointer,                              \
                           PW_TOKENIZER_REPLACE_FORMAT_STRING(__VA_ARGS__)); \
  } while (0)
 
 
#define PW_TOKENIZER_REPLACE_FORMAT_STRING(...) \
  _PW_TOKENIZER_REPLACE_FORMAT_STRING(PW_EMPTY_ARGS(__VA_ARGS__), __VA_ARGS__)
 
#define _PW_TOKENIZER_REPLACE_FORMAT_STRING(empty_args, ...) \
  _PW_CONCAT_2(_PW_TOKENIZER_REPLACE_FORMAT_STRING_, empty_args)(__VA_ARGS__)
 
#define _PW_TOKENIZER_REPLACE_FORMAT_STRING_1() _pw_tokenizer_token, 0u
#define _PW_TOKENIZER_REPLACE_FORMAT_STRING_0(...) \
  _pw_tokenizer_token, PW_TOKENIZER_ARG_TYPES(__VA_ARGS__), __VA_ARGS__
 
#define PW_TOKENIZER_ARG_TYPES(...) \
  PW_DELEGATE_BY_ARG_COUNT(_PW_TOKENIZER_TYPES_, __VA_ARGS__)
 
PW_EXTERN_C_START
 
// These functions encode the tokenized strings. These should not be called
// directly. Instead, use the corresponding PW_TOKENIZE_TO_* macros above.
void _pw_tokenizer_ToBuffer(void* buffer,
                            size_t* buffer_size_bytes,  // input and output arg
                            pw_tokenizer_Token token,
                            pw_tokenizer_ArgTypes types,
                            ...);
 
// This empty function allows the compiler to check the format string.
static inline void pw_tokenizer_CheckFormatString(const char* format, ...)
    PW_PRINTF_FORMAT(1, 2);
 
static inline void pw_tokenizer_CheckFormatString(const char* format, ...) {
  (void)format;
}
 
PW_EXTERN_C_END
 
// clang-format off
#define PW_TOKENIZE_FORMAT_STRING(domain, mask, format, ...)                   \
  static_assert(                                                               \
      PW_FUNCTION_ARG_COUNT(__VA_ARGS__) <= PW_TOKENIZER_MAX_SUPPORTED_ARGS,   \
      "Tokenized strings cannot have more than "                               \
      PW_STRINGIFY(PW_TOKENIZER_MAX_SUPPORTED_ARGS) " arguments; "             \
      PW_STRINGIFY(PW_FUNCTION_ARG_COUNT(__VA_ARGS__))                         \
      " arguments were used for " #format " (" #__VA_ARGS__ ")");              \
  PW_TOKENIZE_FORMAT_STRING_ANY_ARG_COUNT(domain, mask, format, __VA_ARGS__)
// clang-format on
 
#define PW_TOKENIZE_FORMAT_STRING_ANY_ARG_COUNT(domain, mask, format, ...)     \
  if (0) { /* Do not execute to prevent double evaluation of the arguments. */ \
    pw_tokenizer_CheckFormatString(format PW_COMMA_ARGS(__VA_ARGS__));         \
  }                                                                            \
                                                                               \
  _PW_TOKENIZE_VALIDATE_FORMAT_STRING(format);                                 \
                                                                               \
  /* Tokenize the string to a pw_tokenizer_Token at compile time. */           \
  static _PW_TOKENIZER_CONST pw_tokenizer_Token _pw_tokenizer_token =          \
      _PW_TOKENIZER_MASK_TOKEN(mask, format);                                  \
                                                                               \
  PW_TOKENIZER_DEFINE_TOKEN(_pw_tokenizer_token, domain, format)
 
// Creates unique names to use for tokenized string entries and linker sections.
#define _PW_TOKENIZER_UNIQUE(prefix) PW_CONCAT(prefix, __LINE__, _, __COUNTER__)
 
#ifdef __cplusplus
 
#define _PW_TOKENIZER_CONST constexpr
 
#define PW_TOKENIZER_DEFINE_TOKEN(token, domain, string)                       \
  static_assert(::pw::tokenizer::internal::ValidDomain(domain),                \
                "pw_tokenizer domains may only contain alphanumeric "          \
                "characters, underscore, or colon, and cannot start with a "   \
                "number; space characters are ignored");                       \
  alignas(1) static constexpr auto _PW_TOKENIZER_SECTION _PW_TOKENIZER_UNIQUE( \
      _pw_tokenizer_string_entry_) =                                           \
      ::pw::tokenizer::internal::MakeEntry(token, domain, string)
 
// Validates the format string provided to PW_TOKENIZE_FORMAT_STRING and
// friends.
#define _PW_TOKENIZE_VALIDATE_FORMAT_STRING(format)                     \
  do {                                                                  \
    static_assert(!::pw::tokenizer::internal::Contains(format, "%.*s"), \
                  "The %.*s specifier is not supported."                \
                  " See https://pwbug.dev/408040194");                  \
  } while (0)
 
namespace pw::tokenizer {
 
using Token = ::pw_tokenizer_Token;
inline constexpr const char* kDefaultDomain = PW_TOKENIZER_DEFAULT_DOMAIN;
 
namespace internal {
 
constexpr bool Contains(const char* haystack, const char* needle) {
  std::string_view haystack_view(haystack);
  return haystack_view.find(needle) != std::string_view::npos;
}
 
}  // namespace internal
}  // namespace pw::tokenizer
 
#else
 
#define _PW_TOKENIZER_CONST const
#define _PW_ALIGNAS(alignment) __attribute__((aligned(alignment)))
 
#define PW_TOKENIZER_DEFINE_TOKEN(token, domain, string) \
  _PW_ALIGNAS(1) static const _PW_TOKENIZER_STRING_ENTRY(token, domain, string)
 
// There is no way to do this in C.
#define _PW_TOKENIZE_VALIDATE_FORMAT_STRING(format) \
  do {                                              \
  } while (0)
 
#endif  // __cplusplus
 
// _PW_TOKENIZER_SECTION places the tokenized strings in a special .pw_tokenizer
// linker section. Host-side decoding tools read the strings and tokens from
// this section to build a database of tokenized strings.
//
// This section should be declared as type INFO so that it is excluded from the
// final binary. To declare the section, as well as the .pw_tokenizer.info
// metadata section, add the following to the linker script's SECTIONS command:
//
//   .pw_tokenizer.info 0x0 (INFO) :
//   {
//     KEEP(*(.pw_tokenizer.info))
//   }
//
//   .pw_tokenizer.entries 0x0 (INFO) :
//   {
//     KEEP(*(.pw_tokenizer.entries.*))
//   }
//
// A linker script snippet that provides these sections is provided in the file
// pw_tokenizer_linker_sections.ld. This file may be directly included into
// existing linker scripts.
//
// The tokenized string sections can also be managed without linker script
// modifications, though this is not recommended. The section can be extracted
// and removed from the ELF with objcopy:
//
//   objcopy --only-section .pw_tokenizer.* <ORIGINAL_ELF> <OUTPUT_ELF>
//   objcopy --remove-section .pw_tokenizer.* <ORIGINAL_ELF>
//
// OUTPUT_ELF will be an ELF with only the tokenized strings, and the original
// ELF file will have the sections removed.
//
// Without the above linker script modifications, the section garbage collection
// option (--gc-sections) removes the tokenized string sections. To avoid
// editing the target linker script, a separate metadata ELF can be linked
// without --gc-sections to preserve the tokenized data.
//
// pw_tokenizer is intended for use with ELF files only. Mach-O files (macOS
// executables) do not support section names longer than 16 characters, so a
// short, unused section name is used on macOS.
#ifdef __APPLE__
#define _PW_TOKENIZER_SECTION \
  PW_KEEP_IN_SECTION(PW_STRINGIFY(_PW_TOKENIZER_UNIQUE(.pw.)))
#else
#define _PW_TOKENIZER_SECTION \
  PW_KEEP_IN_SECTION(PW_STRINGIFY(_PW_TOKENIZER_UNIQUE(.pw_tokenizer.entries.)))
#endif  // __APPLE__