aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/libs/sparsehash/NEWS
blob: 4af929c394954c4e09a02baae685396136c095cd (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
== 12 October 2015 ==

Various small fixes to ensure compilation on modern compilers and operating 
systems. Tagged as 2.0.3

== 23 February 2012 ==

A backwards incompatibility arose from flattening the include headers
structure for the <google> folder. 

This is now fixed in 2.0.2. You only need to upgrade if you had previously
included files from the <google/sparsehash> folder.

== 1 February 2012 ==

A minor bug related to the namespace switch from google to sparsehash
stopped the build from working when perftools is also installed.

This is now fixed in 2.0.1. You only need to upgrade if you have perftools
installed.  

== 31 January 2012 ==

I've just released sparsehash 2.0.

The `google-sparsehash` project has been renamed to `sparsehash`.  I
(csilvers) am stepping down as maintainer, to be replaced by the team
of Donovan Hide and Geoff Pike.  Welcome to the team, Donovan and
Geoff!  Donovan has been an active contributor to sparsehash bug
reports and discussions in the past, and Geoff has been closely
involved with sparsehash inside Google (in addition to writing the
[http://code.google.com/p/cityhash CityHash hash function]).  The two
of them together should be a formidable force.  For good.

I bumped the major version number up to 2 to reflect the new community
ownership of the project.  All the
[http://sparsehash.googlecode.com/svn/tags/sparsehash-2.0/ChangeLog changes]
are related to the renaming.

The only functional change from sparsehash 1.12 is that I've renamed
the `google/` include-directory to be `sparsehash/` instead.  New code
should `#include <sparsehash/sparse_hash_map>`/etc.  I've kept the old
names around as forwarding headers to the new, so `#include
<google/sparse_hash_map>` will continue to work.

Note that the classes and functions remain in the `google` C++
namespace (I didn't change that to `sparsehash` as well); I think
that's a trickier transition, and can happen in a future release.


=== 18 January 2011 ===

The `google-sparsehash` Google Code page has been renamed to
`sparsehash`, in preparation for the project being renamed to
`sparsehash`.  In the coming weeks, I'll be stepping down as
maintainer for the sparsehash project, and as part of that Google is
relinquishing ownership of the project; it will now be entirely
community run.  The name change reflects that shift.


=== 20 December 2011 ===

I've just released sparsehash 1.12.  This release features improved
I/O (serialization) support.  Support is finally added to serialize
and unserialize `dense_hash_map`/`set`, paralleling the existing code
for `sparse_hash_map`/`set`.  In addition, the serialization API has
gotten simpler, with a single `serialize()` method to write to disk,
and an `unserialize()` method to read from disk.  Finally, support has
gotten more generic, with built-in support for both C `FILE*`s and C++
streams, and an extension mechanism to support arbitrary sources and
sinks.

There are also more minor changes, including minor bugfixes, an
improved deleted-key test, and a minor addition to the `sparsetable`
API.  See the [http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.12/ChangeLog ChangeLog]
for full details.

=== 23 June 2011 ===

I've just released sparsehash 1.11.  The major user-visible change is
that the default behavior is improved -- using the hash_map/set is
faster -- for hashtables where the key is a pointer.  We now notice
that case and ignore the low 2-3 bits (which are almost always 0 for
pointers) when hashing.

Another user-visible change is we've removed the tests for whether the
STL (vector, pair, etc) is defined in the 'std' namespace.  gcc 2.95
is the most recent compiler I know of to put STL types and functions
in the global namespace.  If you need to use such an old compiler, do
not update to the latest sparsehash release.

We've also changed the internal tools we use to integrate
Googler-supplied patches to sparsehash into the opensource release.
These new tools should result in more frequent updates with better
change descriptions.  They will also result in future ChangeLog
entries being much more verbose (for better or for worse).

A full list of changes is described in 
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.11/ChangeLog ChangeLog].

=== 21 January 2011 ===

I've just released sparsehash 1.10.  This fixes a performance
regression in sparsehash 1.8, where sparse_hash_map would copy
hashtable keys by value even when the key was explicitly a reference.
It also fixes compiler warnings from MSVC 10, which uses some c++0x
features that did not interact well with sparsehash.

There is no reason to upgrade unless you use references for your
hashtable keys, or compile with MSVC 10.  A full list of changes is
described in
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.10/ChangeLog ChangeLog].


=== 24 September 2010 ===

I've just released sparsehash 1.9.  This fixes a size regression in
sparsehash 1.8, where the new allocator would take up space in
`sparse_hash_map`, doubling the sparse_hash_map overhead (from 1-2
bits per bucket to 3 or so).  All users are encouraged to upgrade.

This change also marks enums as being Plain Old Data, which can speed
up hashtables with enum keys and/or values.  A full list of changes is
described in
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.9/ChangeLog ChangeLog].

=== 29 July 2010 ===

I've just released sparsehash 1.8.  This includes improved support for
`Allocator`, including supporting the allocator constructor arg and
`get_allocator()` access method.

To work around a bug in gcc 4.0.x, I've renamed the static variables
`HT_OCCUPANCY_FLT` and `HT_SHRINK_FLT` to `HT_OCCUPANCY_PCT` and
`HT_SHRINK_PCT`, and changed their type from float to int.  This
should not be a user-visible change, since these variables are only
used in the internal hashtable classes (sparsehash clients should use
`max_load_factor()` and `min_load_factor()` instead of modifying these
static variables), but if you do access these constants, you will need
to change your code.

Internally, the biggest change is a revamp of the test suite.  It now
has more complete coverage, and a more capable timing tester.  There
are other, more minor changes as well.  A full list of changes is
described in the
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.8/ChangeLog ChangeLog].

=== 31 March 2010 ===

I've just released sparsehash 1.7.  The major news here is the
addition of `Allocator` support.  Previously, these hashtable classes
would just ignore the `Allocator` template parameter.  They now
respect it, and even inherit `size_type`, `pointer`, etc. from the
allocator class.  By default, they use a special allocator we provide
that uses libc `malloc` and `free` to allocate.  The hash classes
notice when this special allocator is being used, and use `realloc`
when it can.  This means that the default allocator is significantly
faster than custom allocators are likely to be (since realloc-like
functionality is not supported by STL allocators).

There are a few more minor changes as well.  A full list of changes is
described in the
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.7/ChangeLog ChangeLog].

=== 11 January 2010 ===

I've just released sparsehash 1.6.  The API has widened a bit with the
addition of `deleted_key()` and `empty_key()`, which let you query
what values these keys have.  A few rather obscure bugs have been
fixed (such as an error when copying one hashtable into another when
the empty_keys differ).  A full list of changes is described in the
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.6/ChangeLog ChangeLog].

=== 9 May 2009 ===

I've just released sparsehash 1.5.1.  Hot on the heels of sparsehash
1.5, this release fixes a longstanding bug in the sparsehash code,
where `equal_range` would always return an empty range.  It now works
as documented.  All sparsehash users are encouraged to upgrade.

=== 7 May 2009 ===

I've just released sparsehash 1.5.  This release introduces tr1
compatibility: I've added `rehash`, `begin(i)`, and other methods that
are expected to be part of the `unordered_map` API once `tr1` in
introduced.  This allows `sparse_hash_map`, `dense_hash_map`,
`sparse_hash_set`, and `dense_hash_set` to be (almost) drop-in
replacements for `unordered_map` and `unordered_set`.

There is no need to upgrade unless you need this functionality, or
need one of the other, more minor, changes described in the
[http://google-sparsehash.googlecode.com/svn/tags/sparsehash-1.5/ChangeLog ChangeLog].