diff options
Diffstat (limited to 'include/ruby/internal/core/rstring.h')
| -rw-r--r-- | include/ruby/internal/core/rstring.h | 399 |
1 files changed, 338 insertions, 61 deletions
diff --git a/include/ruby/internal/core/rstring.h b/include/ruby/internal/core/rstring.h index d073da1d2c..0bca74e688 100644 --- a/include/ruby/internal/core/rstring.h +++ b/include/ruby/internal/core/rstring.h @@ -17,7 +17,7 @@ * recursively included from extension libraries written in C++. * Do not expect for instance `__VA_ARGS__` is always available. * We assume C99 for ruby itself but we don't assume languages of - * extension libraries. They could be written in C++98. + * extension libraries. They could be written in C++98. * @brief Defines struct ::RString. */ #include "ruby/internal/config.h" @@ -32,84 +32,341 @@ #include "ruby/internal/warning_push.h" #include "ruby/assert.h" +/** + * Convenient casting macro. + * + * @param obj An object, which is in fact an ::RString. + * @return The passed object casted to ::RString. + */ #define RSTRING(obj) RBIMPL_CAST((struct RString *)(obj)) -#define RSTRING_NOEMBED RSTRING_NOEMBED -#define RSTRING_EMBED_LEN_MASK RSTRING_EMBED_LEN_MASK -#define RSTRING_EMBED_LEN_SHIFT RSTRING_EMBED_LEN_SHIFT -#define RSTRING_EMBED_LEN_MAX RSTRING_EMBED_LEN_MAX -#define RSTRING_FSTR RSTRING_FSTR /** @cond INTERNAL_MACRO */ -#define RSTRING_EMBED_LEN RSTRING_EMBED_LEN +#define RSTRING_NOEMBED RSTRING_NOEMBED +#define RSTRING_FSTR RSTRING_FSTR #define RSTRING_LEN RSTRING_LEN #define RSTRING_LENINT RSTRING_LENINT #define RSTRING_PTR RSTRING_PTR #define RSTRING_END RSTRING_END /** @endcond */ +/** + * @name Conversion of Ruby strings into C's + * + * @{ + */ + +/** + * Ensures that the parameter object is a String. This is done by calling its + * `to_str` method. + * + * @param[in,out] v Arbitrary Ruby object. + * @exception rb_eTypeError No implicit conversion defined. + * @post `v` is a String. + */ #define StringValue(v) rb_string_value(&(v)) + +/** + * Identical to #StringValue, except it returns a `char*`. + * + * @param[in,out] v Arbitrary Ruby object. + * @exception rb_eTypeError No implicit conversion defined. + * @return Converted Ruby string's backend C string. + * @post `v` is a String. + */ #define StringValuePtr(v) rb_string_value_ptr(&(v)) + +/** + * Identical to #StringValuePtr, except it additionally checks for the contents + * for viability as a C string. Ruby can accept wider range of contents as + * strings, compared to C. This function is to check that. + * + * @param[in,out] v Arbitrary Ruby object. + * @exception rb_eTypeError No implicit conversion defined. + * @exception rb_eArgError String is not C-compatible. + * @return Converted Ruby string's backend C string. + * @post `v` is a String. + */ #define StringValueCStr(v) rb_string_value_cstr(&(v)) + +/** + * @private + * + * @deprecated This macro once was a thing in the old days, but makes no sense + * any longer today. Exists here for backwards compatibility + * only. You can safely forget about it. + */ #define SafeStringValue(v) StringValue(v) + +/** + * Identical to #StringValue, except it additionally converts the string's + * encoding to default external encoding. Ruby has a concept called encodings. + * A string can have different encoding than the environment expects. Someone + * has to make sure its contents be converted to something suitable. This is + * that routine. Call it when necessary. + * + * @param[in,out] v Arbitrary Ruby object. + * @exception rb_eTypeError No implicit conversion defined. + * @return Converted Ruby string's backend C string. + * @post `v` is a String. + * + * @internal + * + * Not sure but it seems this macro does not raise on encoding + * incompatibilities? Doesn't sound right to @shyouhei. + */ #define ExportStringValue(v) do { \ StringValue(v); \ (v) = rb_str_export(v); \ } while (0) +/** @} */ + +/** + * @private + * + * Bits that you can set to ::RBasic::flags. + * + * @warning These enums are not the only bits we use for strings. + * + * @internal + * + * Actually all bits through FL_USER1 to FL_USER19 are used for strings. Why + * only this tiny part of them are made public here? @shyouhei can find no + * reason. + */ enum ruby_rstring_flags { + + /** + * This flag has something to do with memory footprint. If the string is + * short enough, ruby tries to be creative to abuse padding bits of struct + * ::RString for storing contents. If this flag is set that string does + * _not_ do that, to resort to good old fashioned external allocation + * strategy instead. + * + * @warning This bit has to be considered read-only. Setting/clearing + * this bit without corresponding fix up must cause immediate + * SEGV. Also, internal structures of a string change + * dynamically and transparently throughout of its lifetime. + * Don't assume it being persistent. + * + * @internal + * + * 3rd parties must not be aware that there even is more than one way to + * store a string. Might better be hidden. + */ RSTRING_NOEMBED = RUBY_FL_USER1, - RSTRING_EMBED_LEN_MASK = RUBY_FL_USER2 | RUBY_FL_USER3 | RUBY_FL_USER4 | - RUBY_FL_USER5 | RUBY_FL_USER6, + /* Actually, string encodings are also encoded into the flags, using * remaining bits.*/ - RSTRING_FSTR = RUBY_FL_USER17 -}; -enum ruby_rstring_consts { - RSTRING_EMBED_LEN_SHIFT = RUBY_FL_USHIFT + 2, - RSTRING_EMBED_LEN_MAX = RBIMPL_EMBED_LEN_MAX_OF(char) - 1 + /** + * This flag has something to do with infamous "f"string. What is a + * fstring? Well it is a special subkind of strings that is immutable, + * deduped globally, and managed by our GC. It is much like a Symbol (in + * fact Symbols are dynamic these days and are backended using fstrings). + * This concept has been silently introduced at some point in 2.x era. + * Since then it gained wider acceptance in the core. But extension + * libraries could not know that until very recently. Strings of this flag + * live in a special Limbo deep inside of the interpreter. Never try to + * manipulate it by hand. + * + * @internal + * + * Fstrings are not the only variant strings that we implement today. + * Other things are behind-the-scene. This is the only one that is visible + * from extension library. There is no clear reason why it has to be. + * Given there are more "polite" ways to create fstrings, it seems this bit + * need not be exposed to extension libraries. Might better be hidden. + */ + RSTRING_FSTR = RUBY_FL_USER17 }; +/** + * Ruby's String. A string in ruby conceptually has these information: + * + * - Encoding of the string. + * - Length of the string. + * - Contents of the string. + * + * It is worth noting that a string is _not_ an array of characters in ruby. + * It has never been. In 1.x a string was an array of integers. Since 2.x a + * string is no longer an array of anything. A string is a string -- just like + * a Time is not an integer. + */ struct RString { + + /** Basic part, including flags and class. */ struct RBasic basic; + + /** + * Length of the string, not including terminating NUL character. + * + * @note This is in bytes. + */ + long len; + + /** String's specific fields. */ union { + + /** + * Strings that use separated memory region for contents use this + * pattern. + */ struct { - long len; + /** + * Pointer to the contents of the string. In the old days each + * string had dedicated memory regions. That is no longer true + * today, but there still are strings of such properties. This + * field could be used to point such things. + */ char *ptr; + + /** Auxiliary info. */ union { + + /** + * Capacity of `*ptr`. A continuous memory region of at least + * `capa` bytes is expected to exist at `*ptr`. This can be + * bigger than `len`. + */ long capa; + + /** + * Parent of the string. Nowadays strings can share their + * contents each other, constructing gigantic nest of objects. + * This situation is called "shared", and this is the field to + * control such properties. + */ VALUE shared; } aux; } heap; - char ary[RSTRING_EMBED_LEN_MAX + 1]; + + /** Embedded contents. */ + struct { + /* This is a length 1 array because: + * 1. GCC has a bug that does not optimize C flexible array members + * (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102452) + * 2. Zero length arrays are not supported by all compilers + */ + char ary[1]; + } embed; } as; }; RBIMPL_SYMBOL_EXPORT_BEGIN() -VALUE rb_str_to_str(VALUE); -VALUE rb_string_value(volatile VALUE*); -char *rb_string_value_ptr(volatile VALUE*); -char *rb_string_value_cstr(volatile VALUE*); -VALUE rb_str_export(VALUE); -VALUE rb_str_export_locale(VALUE); +/** + * Identical to rb_check_string_type(), except it raises exceptions in case of + * conversion failures. + * + * @param[in] obj Target object. + * @exception rb_eTypeError No implicit conversion to String. + * @return Return value of `obj.to_str`. + * @see rb_io_get_io + * @see rb_ary_to_ary + */ +VALUE rb_str_to_str(VALUE obj); + +/** + * Identical to rb_str_to_str(), except it fills the passed pointer with the + * converted object. + * + * @param[in,out] ptr Pointer to a variable of target object. + * @exception rb_eTypeError No implicit conversion to String. + * @return Return value of `obj.to_str`. + * @post `*ptr` is the return value. + */ +VALUE rb_string_value(volatile VALUE *ptr); + +/** + * Identical to rb_str_to_str(), except it returns the converted string's + * backend memory region. + * + * @param[in,out] ptr Pointer to a variable of target object. + * @exception rb_eTypeError No implicit conversion to String. + * @post `*ptr` is the return value of `obj.to_str`. + * @return Pointer to the contents of the return value. + */ +char *rb_string_value_ptr(volatile VALUE *ptr); + +/** + * Identical to rb_string_value_ptr(), except it additionally checks for the + * contents for viability as a C string. Ruby can accept wider range of + * contents as strings, compared to C. This function is to check that. + * + * @param[in,out] ptr Pointer to a variable of target object. + * @exception rb_eTypeError No implicit conversion to String. + * @exception rb_eArgError String is not C-compatible. + * @post `*ptr` is the return value of `obj.to_str`. + * @return Pointer to the contents of the return value. + */ +char *rb_string_value_cstr(volatile VALUE *ptr); + +/** + * Identical to rb_str_to_str(), except it additionally converts the string + * into default external encoding. Ruby has a concept called encodings. A + * string can have different encoding than the environment expects. Someone + * has to make sure its contents be converted to something suitable. This is + * that routine. Call it when necessary. + * + * @param[in] obj Target object. + * @exception rb_eTypeError No implicit conversion to String. + * @return Converted ruby string of default external encoding. + */ +VALUE rb_str_export(VALUE obj); + +/** + * Identical to rb_str_export(), except it converts into the locale encoding + * instead. + * + * @param[in] obj Target object. + * @exception rb_eTypeError No implicit conversion to String. + * @return Converted ruby string of locale encoding. + */ +VALUE rb_str_export_locale(VALUE obj); RBIMPL_ATTR_ERROR(("rb_check_safe_str() and Check_SafeStr() are obsolete; use StringValue() instead")) +/** + * @private + * + * @deprecated This function once was a thing in the old days, but makes no + * sense any longer today. Exists here for backwards + * compatibility only. You can safely forget about it. + */ void rb_check_safe_str(VALUE); + +/** + * @private + * + * @deprecated This macro once was a thing in the old days, but makes no sense + * any longer today. Exists here for backwards compatibility + * only. You can safely forget about it. + */ #define Check_SafeStr(v) rb_check_safe_str(RBIMPL_CAST((VALUE)(v))) + +/** + * @private + * + * Prints diagnostic message to stderr when RSTRING_PTR or RSTRING_END + * is NULL. + * + * @param[in] func The function name where encountered NULL pointer. + */ +void rb_debug_rstring_null_ptr(const char *func); RBIMPL_SYMBOL_EXPORT_END() RBIMPL_ATTR_PURE_UNLESS_DEBUG() RBIMPL_ATTR_ARTIFICIAL() +/** + * Queries the length of the string. + * + * @param[in] str String in question. + * @return Its length, in bytes. + * @pre `str` must be an instance of ::RString. + */ static inline long -RSTRING_EMBED_LEN(VALUE str) +RSTRING_LEN(VALUE str) { - RBIMPL_ASSERT_TYPE(str, RUBY_T_STRING); - RBIMPL_ASSERT_OR_ASSUME(! RB_FL_ANY_RAW(str, RSTRING_NOEMBED)); - - VALUE f = RBASIC(str)->flags; - f &= RSTRING_EMBED_LEN_MASK; - f >>= RSTRING_EMBED_LEN_SHIFT; - return RBIMPL_CAST((long)f); + return RSTRING(str)->len; } RBIMPL_WARNING_PUSH() @@ -119,6 +376,15 @@ RBIMPL_WARNING_IGNORED(413) RBIMPL_ATTR_PURE_UNLESS_DEBUG() RBIMPL_ATTR_ARTIFICIAL() +/** + * @private + * + * "Expands" an embedded string into an ordinal one. This is a function that + * returns aggregated type. The returned struct always has its `as.heap.len` + * an `as.heap.ptr` fields set appropriately. + * + * This is an implementation detail that 3rd parties should never bother. + */ static inline struct RString rbimpl_rstring_getmem(VALUE str) { @@ -130,82 +396,93 @@ rbimpl_rstring_getmem(VALUE str) else { /* Expecting compilers to optimize this on-stack struct away. */ struct RString retval; - retval.as.heap.len = RSTRING_EMBED_LEN(str); - retval.as.heap.ptr = RSTRING(str)->as.ary; + retval.len = RSTRING_LEN(str); + retval.as.heap.ptr = RSTRING(str)->as.embed.ary; return retval; } } RBIMPL_WARNING_POP() -RBIMPL_ATTR_PURE_UNLESS_DEBUG() -RBIMPL_ATTR_ARTIFICIAL() -static inline long -RSTRING_LEN(VALUE str) -{ - return rbimpl_rstring_getmem(str).as.heap.len; -} - RBIMPL_ATTR_ARTIFICIAL() +/** + * Queries the contents pointer of the string. + * + * @param[in] str String in question. + * @return Pointer to its contents. + * @pre `str` must be an instance of ::RString. + */ static inline char * RSTRING_PTR(VALUE str) { char *ptr = rbimpl_rstring_getmem(str).as.heap.ptr; - if (RB_UNLIKELY(! ptr)) { + if (RUBY_DEBUG && RB_UNLIKELY(! ptr)) { /* :BEWARE: @shyouhei thinks that currently, there are rooms for this - * function to return NULL. In the 20th century that was a pointless - * concern. However struct RString can hold fake strings nowadays. It - * seems no check against NULL are exercised around handling of them - * (one of such usages is located in marshal.c, which scares - * @shyouhei). Better check here for maximum safety. + * function to return NULL. Better check here for maximum safety. * * Also, this is not rb_warn() because RSTRING_PTR() can be called * during GC (see what obj_info() does). rb_warn() needs to allocate * Ruby objects. That is not possible at this moment. */ - fprintf(stderr, "%s\n", - "RSTRING_PTR is returning NULL!! " - "SIGSEGV is highly expected to follow immediately. " - "If you could reproduce, attach your debugger here, " - "and look at the passed string." - ); + rb_debug_rstring_null_ptr("RSTRING_PTR"); } return ptr; } RBIMPL_ATTR_ARTIFICIAL() +/** + * Queries the end of the contents pointer of the string. + * + * @param[in] str String in question. + * @return Pointer to its end of contents. + * @pre `str` must be an instance of ::RString. + */ static inline char * RSTRING_END(VALUE str) { struct RString buf = rbimpl_rstring_getmem(str); - if (RB_UNLIKELY(! buf.as.heap.ptr)) { + if (RUBY_DEBUG && RB_UNLIKELY(! buf.as.heap.ptr)) { /* Ditto. */ - fprintf(stderr, "%s\n", - "RSTRING_END is returning NULL!! " - "SIGSEGV is highly expected to follow immediately. " - "If you could reproduce, attach your debugger here, " - "and look at the passed string." - ); + rb_debug_rstring_null_ptr("RSTRING_END"); } - return &buf.as.heap.ptr[buf.as.heap.len]; + return &buf.as.heap.ptr[buf.len]; } RBIMPL_ATTR_ARTIFICIAL() +/** + * Identical to RSTRING_LEN(), except it differs for the return type. + * + * @param[in] str String in question. + * @exception rb_eRangeError Too long. + * @return Its length, in bytes. + * @pre `str` must be an instance of ::RString. + * + * @internal + * + * This API seems redundant but has actual usages. + */ static inline int RSTRING_LENINT(VALUE str) { return rb_long2int(RSTRING_LEN(str)); } +/** + * Convenient macro to obtain the contents and length at once. + * + * @param str String in question. + * @param ptrvar Variable where its contents is stored. + * @param lenvar Variable where its length is stored. + */ #ifdef HAVE_STMT_AND_DECL_IN_EXPR # define RSTRING_GETMEM(str, ptrvar, lenvar) \ __extension__ ({ \ struct RString rbimpl_str = rbimpl_rstring_getmem(str); \ (ptrvar) = rbimpl_str.as.heap.ptr; \ - (lenvar) = rbimpl_str.as.heap.len; \ + (lenvar) = rbimpl_str.len; \ }) #else # define RSTRING_GETMEM(str, ptrvar, lenvar) \ |