1 // stringpool.h -- a string pool for gold -*- C++ -*-
3 // Copyright 2006, 2007 Free Software Foundation, Inc.
4 // Written by Ian Lance Taylor <iant@google.com>.
6 // This file is part of gold.
8 // This program is free software; you can redistribute it and/or modify
9 // it under the terms of the GNU General Public License as published by
10 // the Free Software Foundation; either version 3 of the License, or
11 // (at your option) any later version.
13 // This program is distributed in the hope that it will be useful,
14 // but WITHOUT ANY WARRANTY; without even the implied warranty of
15 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 // GNU General Public License for more details.
18 // You should have received a copy of the GNU General Public License
19 // along with this program; if not, write to the Free Software
20 // Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
21 // MA 02110-1301, USA.
26 #ifndef GOLD_STRINGPOOL_H
27 #define GOLD_STRINGPOOL_H
34 // A Stringpool is a pool of unique strings. It provides the
35 // following features:
37 // Every string in the pool is unique. Thus, if you have two strings
38 // in the Stringpool, you can compare them for equality by using
39 // pointer comparison rather than string comparison.
41 // There is a key associated with every string in the pool. If you
42 // add strings to the Stringpool in the same order, then the key for
43 // each string will always be the same for any run of the linker.
44 // This is not true of the string pointers themselves, as they may
45 // change due to address space randomization. Some parts of the
46 // linker (e.g., the symbol table) use the key value instead of the
47 // string pointer so that repeated runs of the linker will generate
48 // precisely the same output.
50 // When you add a string to a Stringpool, Stringpool will optionally
51 // make a copy of it. Thus there is no requirement to keep a copy
54 // A Stringpool can be turned into a string table, a sequential series
55 // of null terminated strings. The first string may optionally be a
56 // single zero byte, as required for SHT_STRTAB sections. This
57 // conversion is only permitted after all strings have been added to
58 // the Stringpool. After doing this conversion, you can ask for the
59 // offset of any string in the stringpool in the string table, and you
60 // can write the resulting string table to an output file.
62 // When a Stringpool is turned into a string table, then as an
63 // optimization it will reuse string suffixes to avoid duplicating
64 // strings. That is, given the strings "abc" and "bc", only the
65 // string "abc" will be stored, and "bc" will be represented by an
66 // offset into the middle of the string "abc".
68 // Stringpools are implemented in terms of Stringpool_template, which
69 // is generalized on the type of character used for the strings. Most
70 // uses will want the Stringpool type which uses char. Other cases
71 // are used for merging wide string constants.
73 template<typename Stringpool_char
>
74 class Stringpool_template
77 // The type of a key into the stringpool. As described above, a key
78 // value will always be the same during any run of the linker. Zero
79 // is never a valid key value.
82 // Create a Stringpool.
83 Stringpool_template();
85 ~Stringpool_template();
87 // Clear all the data from the stringpool.
91 // Hint to the stringpool class that you intend to insert n additional
92 // elements. The stringpool class can use this info however it likes;
93 // in practice it will resize its internal hashtables to make room.
95 reserve(unsigned int n
);
97 // Indicate that we should not reserve offset 0 to hold the empty
98 // string when converting the stringpool to a string table. This
99 // should not be called for a proper ELF SHT_STRTAB section.
102 { this->zero_null_
= false; }
104 // Add the string S to the pool. This returns a canonical permanent
105 // pointer to the string in the pool. If COPY is true, the string
106 // is copied into permanent storage. If PKEY is not NULL, this sets
107 // *PKEY to the key for the string.
108 const Stringpool_char
*
109 add(const Stringpool_char
* s
, bool copy
, Key
* pkey
);
111 // Add string S of length LEN characters to the pool. If COPY is
112 // true, S need not be null terminated.
113 const Stringpool_char
*
114 add_with_length(const Stringpool_char
* s
, size_t len
, bool copy
, Key
* pkey
);
116 // If the string S is present in the pool, return the canonical
117 // string pointer. Otherwise, return NULL. If PKEY is not NULL,
118 // set *PKEY to the key.
119 const Stringpool_char
*
120 find(const Stringpool_char
* s
, Key
* pkey
) const;
122 // Turn the stringpool into a string table: determine the offsets of
123 // all the strings. After this is called, no more strings may be
124 // added to the stringpool.
126 set_string_offsets();
128 // Get the offset of the string S in the string table. This returns
129 // the offset in bytes, not in units of Stringpool_char. This may
130 // only be called after set_string_offsets has been called.
132 get_offset(const Stringpool_char
* s
) const;
134 // Get the offset of the string S in the string table.
136 get_offset(const std::basic_string
<Stringpool_char
>& s
) const
137 { return this->get_offset_with_length(s
.c_str(), s
.size()); }
139 // Get the offset of string S, with length LENGTH characters, in the
142 get_offset_with_length(const Stringpool_char
* s
, size_t length
) const;
144 // Get the size of the string table. This returns the number of
145 // bytes, not in units of Stringpool_char.
147 get_strtab_size() const
149 gold_assert(this->strtab_size_
!= 0);
150 return this->strtab_size_
;
153 // Write the string table into the output file at the specified
156 write(Output_file
*, off_t offset
);
158 // Write the string table into the specified buffer, of the
159 // specified size. buffer_size should be at least
160 // get_strtab_size().
162 write_to_buffer(unsigned char* buffer
, section_size_type buffer_size
);
164 // Dump statistical information to stderr.
166 print_stats(const char*) const;
169 Stringpool_template(const Stringpool_template
&);
170 Stringpool_template
& operator=(const Stringpool_template
&);
172 // Return the length of a string in units of Stringpool_char.
174 string_length(const Stringpool_char
*);
176 // Return whether two strings are equal.
178 string_equal(const Stringpool_char
*, const Stringpool_char
*);
180 // Compute a hash code for a string. LENGTH is the length of the
181 // string in characters.
183 string_hash(const Stringpool_char
*, size_t length
);
185 // We store the actual data in a list of these buffers.
188 // Length of data in buffer.
190 // Allocated size of buffer.
198 // Copy a string into the buffers, returning a canonical string.
199 const Stringpool_char
*
200 add_string(const Stringpool_char
*, size_t, Key
*);
202 // Return whether s1 is a suffix of s2.
204 is_suffix(const Stringpool_char
* s1
, size_t len1
,
205 const Stringpool_char
* s2
, size_t len2
);
207 // The hash table key includes the string, the length of the string,
208 // and the hash code for the string. We put the hash code
209 // explicitly into the key so that we can do a find()/insert()
210 // sequence without having to recompute the hash. Computing the
211 // hash code is a significant user of CPU time in the linker.
214 const Stringpool_char
* string
;
215 // Length is in characters, not bytes.
219 // This goes in an STL container, so we need a default
222 : string(NULL
), length(0), hash_code(0)
225 // Note that these constructors are relatively expensive, because
226 // they compute the hash code.
227 explicit Hashkey(const Stringpool_char
* s
)
228 : string(s
), length(string_length(s
)), hash_code(string_hash(s
, length
))
231 Hashkey(const Stringpool_char
* s
, size_t len
)
232 : string(s
), length(len
), hash_code(string_hash(s
, len
))
236 // Hash function. This is trivial, since we have already computed
238 struct Stringpool_hash
241 operator()(const Hashkey
& hk
) const
242 { return hk
.hash_code
; }
245 // Equality comparison function for hash table.
249 operator()(const Hashkey
&, const Hashkey
&) const;
252 // The hash table is a map from strings to a pair of Key and string
253 // table offsets. We only use the offsets if we turn this into an
254 // string table section.
256 typedef std::pair
<Key
, section_offset_type
> Hashval
;
258 typedef Unordered_map
<Hashkey
, Hashval
, Stringpool_hash
,
259 Stringpool_eq
> String_set_type
;
261 // Comparison routine used when sorting into a string table.
263 typedef typename
String_set_type::iterator Stringpool_sort_info
;
265 struct Stringpool_sort_comparison
268 operator()(const Stringpool_sort_info
&, const Stringpool_sort_info
&) const;
271 // List of Stringdata structures.
272 typedef std::list
<Stringdata
*> Stringdata_list
;
274 // Mapping from const char* to namepool entry.
275 String_set_type string_set_
;
277 Stringdata_list strings_
;
278 // Size of string table.
279 section_size_type strtab_size_
;
280 // Next Stringdata index.
281 unsigned int next_index_
;
282 // Next key value for a string we don't copy.
283 int next_uncopied_key_
;
284 // Whether to reserve offset 0 to hold the null string.
288 // The most common type of Stringpool.
289 typedef Stringpool_template
<char> Stringpool
;
291 } // End namespace gold.
293 #endif // !defined(GOLD_STRINGPOOL_H)