Inderjeet Singh
c25278b4d6
Created an alpha package that holds experimental feature.
...
Added support for JsonPostDeserializer that allows you to invoke postDeserialize methods on an Gson deserialized object.
2012-10-11 03:20:36 +00:00
Joel Leitch
14f16e2d0c
Adding Red-Black Tree implementation and tying it into the Gson bindings.
2012-10-11 03:15:49 +00:00
Jesse Wilson
2fef83799d
Optimizations and bug fixes for LinkedHashTreeMap.
...
The most interesting optimization is to replace ArrayDeque with a manual linked list that reuses the nodes 'parent' field. These optimizations save about 20%.
2012-09-17 03:30:20 +00:00
Jesse Wilson
aceadaecf1
Add a hashing layer to LinkedTreeMap. Instead of having 1 root node, the class now has several root nodes, one for each hash bucket in a hash table.
...
Compared to LinkedTreeMap, this is slower for small (size=5) maps: 124% slower to get() and 33% slower to create and populate. It's a win for large (size=500) maps: 46% faster to get() but 8% slower to create and populate. And it's a big win for very large (size=50,000) maps: 81% faster to get() and 46% faster to create and populate.
http://microbenchmarks.appspot.com/run/limpbizkit@gmail.com/com.google.common.collect.MapBenchmark
I'm going to follow this up with some simple optimizations: caching local fields and simplifying access. That should narrow the performance gap.
2012-09-17 00:19:44 +00:00
Jesse Wilson
a0493b9732
New code that can split an AVL tree into two AVL trees.
...
This is in preparation for a new feature where LinkedTreeMap will have multiple roots, each in its own hash bucket.
2012-09-15 06:13:33 +00:00
Jesse Wilson
01bd0d92e2
Tests for LinkedTreeMap with incompatible keys.
2012-09-12 05:06:48 +00:00
Jesse Wilson
a6ab854302
Fix a bug where we were unlinking nodes that shouldn't have been unlinked.
...
Found by Guava's awesome collections test suite!
2012-09-12 04:41:58 +00:00
Jesse Wilson
93e38901df
Draft of LinkedTreeMap. Its ordered like a LinkedHashMap but it doesn't do any hashing for DoS resistance.
...
Not yet adopted in our code.
Known critical bugs:
- throws ClassCastException when get() is called with a non-comparable key
- throws NullPointerException on get(null)
2012-09-10 20:04:38 +00:00
Jesse Wilson
084047d80b
Reintroduce string pooling in JsonReader.
...
This makes Hotspot slower. From my before/after measurements using ParseBenchmark, times in microseconds:
TWEETS: 350 -> 370 (+6%)
READER_SHORT: 77 -> 76 (-1%)
READER_LONG: 870 -> 940 (+8%)
But it makes Dalvik faster by a greater margin. These before/after measurements use times in milliseconds:
TWEETS: 25 -> 20 (-20%)
READER_SHORT: 5.6 -> 4.7 (-16%)
READER_LONG: 52 -> 47 (-10%)
It's a net win because we're saving a greater fraction of time, and because we're helping the platform that needs the most help. We're paying microseconds on Hotspot to gain milliseconds on Dalvik.
2012-09-10 16:13:33 +00:00
Jesse Wilson
411c5c0b50
Follow up on r1197 and make deepCopy package-private. We don't want to use protected because some of the classes are non-final and protected shows up in the Javadocs.
2012-09-03 23:30:27 +00:00
Inderjeet Singh
a973837dd4
made deepCopy protected to eliminate it from the public API for now.
...
It will be in the subsequent release.
2012-09-03 08:34:37 +00:00
Jesse Wilson
8df7209074
Fix cut & paste issue in JsonToken Javadocs.
...
Fixes issue 463.
2012-09-02 21:29:30 +00:00
Jesse Wilson
b3b919770b
Change number parsing to use one big loop. This changes it to return JsonToken.STRING for very long (>8k digits) numbers.
2012-09-02 20:12:19 +00:00
Jesse Wilson
3920d95fac
rename peekedInteger to peekedLong
2012-09-02 17:46:02 +00:00
Jesse Wilson
8daf3aaeb4
Inline position computation. This is uglier but faster.
2012-08-28 03:52:18 +00:00
Jesse Wilson
46b73632b0
Fix a bug where we weren't reading enough characters when a BOM was encountered.
2012-08-28 01:48:25 +00:00
Jesse Wilson
980796005f
Use a conventional for loop in nextQuotedValue() to make hotspot's job easier.
2012-08-27 04:42:39 +00:00
Jesse Wilson
bdf2cac6d4
Replace switch with if/else when processing whitespace. If/else is faster!
2012-08-27 04:17:29 +00:00
Jesse Wilson
b0a172944a
More number parsing improvements.
2012-08-27 03:17:41 +00:00
Jesse Wilson
448063dde1
Fix a goof in number parsing.
2012-08-27 03:07:20 +00:00
Jesse Wilson
4c2980e6ff
Fix a bug in integer parsing.
2012-08-27 02:34:52 +00:00
Jesse Wilson
c7cb503cdb
Restore fast skips.
...
document api ns linear runtime
TWEETS GSON_STREAM 397568 =========
TWEETS GSON_SKIP 300058 =======
READER_SHORT GSON_STREAM 76632 =
READER_SHORT GSON_SKIP 57796 =
READER_LONG GSON_STREAM 894690 =====================
READER_LONG GSON_SKIP 565114 =============
2012-08-27 01:17:50 +00:00
Jesse Wilson
085856c128
Don't leave the JsonReader in an invalid state if nextInt(), nextDouble() or nextLong() fails. We now save a reference to the string before we parse it, and keep that referenced value if parsing fails.
2012-08-26 22:06:57 +00:00
Jesse Wilson
e7bfd0c97d
Promote JsonReader2 to be the main JsonReader implementation.
2012-08-26 19:34:46 +00:00
Jesse Wilson
553fa6b742
Prepare to replace JsonReader with JsonReader2
2012-08-26 19:33:49 +00:00
Jesse Wilson
b96d2d9837
Support non-execute prefixes.
2012-08-26 19:31:06 +00:00
Jesse Wilson
97cb326ad2
Delete an old version of the rewritten JsonReader
2012-08-26 04:04:22 +00:00
Jesse Wilson
46b4346505
Get JsonReader2 to the point that it's passing most tests.
...
Still missing:
- non-execute prefixes
- rolling back 'pos' when a double fails to parse
- octal prefix failures
2012-08-26 04:02:09 +00:00
Jesse Wilson
c5c65ba626
Add an experimental rewrite of JsonReader.
...
The motivating difference is that JsonReaderV2 tries to read each character at most once. This means that when it reads literals, it also attempts to decode them to a keyword (true/false/null) or a number.
This change also _doesn't_ read strings until demanded to do so. This should permit streaming access to strings down the road.
This code is not yet complete, nor is has it been properly optimized. And the implementation is also quite a mess! It is a work in progress.
2012-08-25 04:31:56 +00:00
Jesse Wilson
9c4b23b39a
JsonObject.deepCopy() for Gson.
2012-08-21 01:19:43 +00:00
Jesse Wilson
1a4f690335
Regretfully enable serialization for StringMap and LazilyParsedNumber.
...
One of our favorite users (my employer!) is stuck in a sad situation where they need to serialize objects returned from Gson; this is a workable escape hatch.
2012-08-15 14:58:26 +00:00
Jesse Wilson
35c13173b0
Switch on ints rather than enums in JsonReader.
...
Using enums was triggering this Android bug:
http://code.google.com/p/android/issues/detail?id=36349
2012-08-14 21:32:18 +00:00
Jesse Wilson
f406d3cf89
Don't fall back when we're already in the slow case.
2012-07-10 23:45:20 +00:00
Jesse Wilson
4aaa4bf20c
StringMap was suffering because the string's hashCode was not cached. Address this by preferring the regular String.hashCode until hash collision problems start to occur.
2012-07-10 18:46:01 +00:00
Inderjeet Singh
b946a229b6
deleted unused field.
2012-07-03 00:28:55 +00:00
Jesse Wilson
4816941f0d
Address code review comments on r1154
2012-07-02 20:32:38 +00:00
Inderjeet Singh
6575cdebca
When EOF is encountered prematurely, Streams.parse() (and JsonParser) now throw JsonSyntaxException.
2012-07-02 18:36:54 +00:00
Inderjeet Singh
582b0a0c9c
Fixed issue 443 by relying on Streams.parse() to return a JsonNull on empty documents and throw a JsonParseException otherwise.
2012-06-30 18:48:11 +00:00
Jesse Wilson
1c7aee40f3
Implement Flushable.
2012-06-30 02:46:26 +00:00
Jesse Wilson
dc4e43bb23
Permit users to define type adapters for primitive types and strings.
...
Also expose an API to get the field naming strategy.
2012-06-30 02:37:49 +00:00
Inderjeet Singh
83e5a4937c
Renamed Gson.getNextAdapter to getDelegateAdapter.
...
Deleted testParameterizedMapSubclassDeserialization which we decided to not fix.
Added simple tests for getDelegateAdapter
2012-04-12 18:49:27 +00:00
Jesse Wilson
3df2db1f16
Don't permit a type adapter for String to be registered.
2012-04-12 18:27:48 +00:00
Jesse Wilson
6d351fea07
Fix a documentation typo.
...
Fixes bug 423.
2012-04-12 18:09:07 +00:00
Jesse Wilson
a991e54157
Make fewer calls to out.write() when serializing strings. On one Android test, this improved serialization time of some documents by 83%.
...
TWEETS
run vm htmlSafe ms linear runtime %
Before app_process true 68.7 ============================== 100%
After app_process true 35.9 =============== 52%
READER_LONG
run vm htmlSafe ms linear runtime %
Before app_process true 439.0 ============================== 100%
After app_process true 74.5 ===== 17%
2012-04-12 13:24:37 +00:00
Jesse Wilson
9be0fd9ecc
Make the BigDecimal and BigInteger type adapters user-overrideable.
2012-03-18 17:55:15 +00:00
Inderjeet Singh
5e3f5a6bbe
Ensuring that the hash is unpredictable and well distributed.
...
We achieve this by using the same algorithm as the Perl version, but this implementation
is being written from scratch for license compliance.
2012-03-15 22:27:55 +00:00
Inderjeet Singh
14ebcc4ead
fixed eclipse warnings
2012-03-11 21:43:38 +00:00
Jesse Wilson
4c629347da
Only support string keys in string map. This rev passed all 655 applicable map tests in the Guava collections test suite.
2012-03-11 17:34:46 +00:00
Jesse Wilson
751c69c655
Support null values in StringMap
2012-03-11 15:19:01 +00:00
Jesse Wilson
ad3489f557
First steps to StringMap, an alternative to LinkedHashmap.
2012-03-11 13:54:41 +00:00
Jesse Wilson
15e7819e9a
Fix testStringEndingInSlash by fixing nextNonWhitespace to always return the character at buffer[pos-1].
2012-02-16 22:49:53 +00:00
Jesse Wilson
2c8bec27d4
Permit multiple top-level values in JsonWriter in lenient mode. Also fix some cases where we don't throw the right thing on a closed JsonWriter.
...
I'd prefer to not support multiple top-level values, but we support it in JsonReader and it's easier to be consistent. Kevin Hayen's patch pointed me in the right direction here, but I needed to do more work to cover some of the edge cases.
Fixes issue 397.
2012-02-12 20:42:16 +00:00
Jesse Wilson
5c978948a0
Remove some dead code.
2012-02-11 20:16:21 +00:00
Jesse Wilson
7b75efd09e
Write some tests prescribed by missing code coverage. I found a bug where our nonexecute prefix code causes a problem.
2012-02-11 20:15:39 +00:00
Jesse Wilson
dd86d63436
Use inner classes for BigDecimal and BigInteger type adapters
2012-02-11 20:14:23 +00:00
Jesse Wilson
bb8dca71c4
Restore ability of instance creators to create collection and map types. We inadvertently lost this in Gson 2.0 and 2.1. Nobody noticed!
2012-01-01 15:46:33 +00:00
Jesse Wilson
6cca23c172
Get GraphAdapterBuilder working for serialization and deserialization using InstanceCreators to get a sneak peek at a value under construction.
2012-01-01 13:42:44 +00:00
Jesse Wilson
d4a1e49e46
Delete some obsolete TODOs
2012-01-01 12:42:48 +00:00
Jesse Wilson
796a381279
Kill GsonInternalAccess. Clients to this were all broken because nobody was ever assigning INSTANCE.
2012-01-01 12:42:20 +00:00
Jesse Wilson
323dfa0af5
Be strict in TypeAdapter's toJson/fromJson methods
2012-01-01 12:36:42 +00:00
Inderjeet Singh
6c78bf5247
made toJson/fromJson/toJsonTree methods public in TypeAdapter.
...
made Gson.getNextAdapter method public.
2011-12-31 08:52:59 +00:00
Inderjeet Singh
498049b304
updated documentation for registerTypeHierarchyAdapter to cover TypeAdapter.
2011-12-31 06:00:28 +00:00
Inderjeet Singh
46d2d79ba7
Added javadocs for type adapter registration through registerTypeAdapter method.
2011-12-31 05:32:14 +00:00
Jesse Wilson
ecdf9150f6
Hide Gson.getNextAdapter() for the current release.
2011-12-31 05:30:40 +00:00
Jesse Wilson
4057b98bab
Implement all but the most difficult part of graph type adapter's deserialization. The catch is we want to return an instance that we don't have yet. It's on the stack, but we don't have a handle to it because it's inside the 'nextTypeAdapter' who is busy populating its fields.
2011-12-30 08:27:24 +00:00
Jesse Wilson
740d03ef0e
Don't call setAccessible(true) on fields we won't be setting or getting.
...
Fixes bug 191.
2011-12-29 07:11:43 +00:00
Jesse Wilson
8d5de3136c
Inline character unescaping. This saves ~10% on the READER_LONG benchmark.
2011-12-25 07:09:46 +00:00
Jesse Wilson
d7fbac0384
Rename TypeAdapter.Factory to TypeAdapterFactory.
2011-12-23 18:27:13 +00:00
Inderjeet Singh
82f18a257f
Implemented code review comments from r1090
2011-12-23 15:52:10 +00:00
Inderjeet Singh
bd937fe7b5
A type adapter for Class that throws an UnsupportedOperationException.
2011-12-22 22:31:43 +00:00
Jesse Wilson
b28e518c7e
Hide toJson/fromJson APIs for the 2.1 release.
2011-12-21 21:30:18 +00:00
Inderjeet Singh
b5ae3c945a
Incorporated code review from r949
2011-12-16 19:10:54 +00:00
Jesse Wilson
a3ca4e1312
Admit to a mistake in InetAddress' type adapter
2011-12-16 19:05:10 +00:00
Jesse Wilson
f24da51ca2
Fix dangling sentence.
2011-12-16 19:00:37 +00:00
Jesse Wilson
c01fc5c935
Fix tests broken by r1078.
2011-12-16 14:12:34 +00:00
Jesse Wilson
2ef7716209
Fix a broken Javadoc link to this.
2011-12-16 05:38:16 +00:00
Jesse Wilson
214234e202
Support @SerializedName on annotations.
...
Fixes issue 347.
2011-12-16 05:32:50 +00:00
Jesse Wilson
b7c3e0067c
Avoid local field accesses in nextString(char). This saves a modest 2%.
2011-12-14 06:07:40 +00:00
Jesse Wilson
d01d39aa26
Use locals instead of fields when figuring out the buffer's offset line and column. This saves about 2% when parsing twitter data.
2011-12-14 05:49:58 +00:00
Jesse Wilson
aa52435951
Apply an ugly optimization to save 5% on pretty printed JSON documents. This uses locals instead of fields in an inner loop to save field reads and writes.
2011-12-14 05:26:29 +00:00
Jesse Wilson
40cd660115
Always provide line and column information when a parse fails.
2011-12-13 23:42:10 +00:00
Jesse Wilson
8e8bf934f9
Inline the nesting stack to save ~20% on JsonReader parsing.
2011-12-13 04:08:48 +00:00
Jesse Wilson
61a549b74d
Don't allocate exceptions in the common case. On one benchmark this improved performance by 20%.
2011-12-13 03:24:35 +00:00
Jesse Wilson
eb2230caf0
Fix nullSafe() to not infinitely recurse on non-null input.
2011-12-06 15:35:52 +00:00
Jesse Wilson
e2e672740a
Fix broken test in registering competing type hierarchy adapters.
...
I actually tried to replicate this test but got an error "type adapters conflict" when I was doing it. I suspect the problem was that I was trying to use 'Object' as the base of my type hierarchy and that class is somehow special.
2011-12-06 15:29:48 +00:00
Inderjeet Singh
b19e187bdd
Minor improvements to TypeAdapter javadocs to promote the use of nullSafe.
2011-12-06 08:32:19 +00:00
Inderjeet Singh
0b734e46e1
Incorporated code review comments from r1061.
...
Made nullSafe() an instance method instead of a static method.
Updated code javadoc to match Guava style.
2011-12-06 08:18:00 +00:00
Jesse Wilson
d5ed0716db
Fix type adapter precedence so that last-registered wins (except for tree type hierarchy adapters, which were always last).
2011-12-06 05:09:18 +00:00
Jesse Wilson
296d843afd
Implement Inder's improvements on the TypeAdapter docs
2011-12-06 04:56:46 +00:00
Inderjeet Singh
91be944022
Added a new API method nullSafe() in TypeAdapter that can be used to avoid boilerplate handling of nulls in a type adapter.
2011-12-05 19:50:49 +00:00
Inderjeet Singh
756131d869
removed Eclipse 3.7 warnings.
2011-12-04 10:24:07 +00:00
Jesse Wilson
f602bce9f5
Nice documentation for TypeAdapter.
2011-12-03 19:46:25 +00:00
Inderjeet Singh
8ee2c24f61
renamed JsonElementReader to JsonTreeReader
2011-12-03 02:37:27 +00:00
Inderjeet Singh
f3c14b4614
Added support for promoteNameToValue for JsonElementReader.
2011-12-03 02:35:46 +00:00
Jesse Wilson
8f8e69a364
Add @since tags.
2011-12-02 23:11:28 +00:00
Jesse Wilson
26ab404599
Cleanup names for TypeAdapters.
2011-12-02 22:57:30 +00:00
Inderjeet Singh
2da01fb183
Deleted deepCopy as GsonBuilder should not be designed as a reusable object.
2011-11-29 07:58:32 +00:00
Jesse Wilson
ec42d600af
Decode JSON literal types eagerly and with our own decoder. Previously we relied on Double.parseDouble() to decode tokens. Since that method is expensive, we deferred calling it unless absolutely necessary. Now we decode the literal type immediately. For efficiency we decode the token right out of the char buffer. This makes things more complicated but it saves many calls to charAt(). It also opens up the possibility to deferring string creation.
2011-11-27 16:50:45 +00:00
Jesse Wilson
852cd72059
Don't allocate a whole bunch of objects each time we deserialize a map key. This saves 20% on bug 375's benchmark.
...
The fix here is quite distasteful; I'm adding an ugly backdoor API for MapTypeAdapterFactory to fiddle with the internal bits of JsonReader. I'd like to revisit this sooner or later, but for now I'll take the speedup.
2011-11-26 15:36:08 +00:00
Jesse Wilson
4c06b01369
Cache all computed type adapters. On one particularly violent test (issue 375) this improves performance by 77%.
2011-11-26 15:30:38 +00:00