128 bit numbers

[Licenser]

See #59

todo:

• bench
• implement 128bit parsing
• remove 128bit as default flag

[Licenser]

Initial perf (just memory layout updates & functions, no parsing).

Impact is measurable but not terrible, nothing to enable by default but as a flag for those who need it it's acceptable performance I think.

0.2.2 (master)

======= simd_json ======== parse|stringify ===== parse|stringify ====
data/canada.json         530 MB/s   420 MB/s   710 MB/s   320 MB/s
data/citm_catalog.json  1130 MB/s   480 MB/s  1550 MB/s   610 MB/s
data/twitter.json        920 MB/s   600 MB/s  1000 MB/s   640 MB/s
data/log.json           1080 MB/s  1080 MB/s  1080 MB/s  1080 MB/s

===== simd_json_tape ===== parse|stringify ===== parse|stringify ====
data/canada.json         770 MB/s
data/citm_catalog.json  1890 MB/s
data/twitter.json       1670 MB/s
data/log.json           2170 MB/s

this

======= simd_json ======== parse|stringify ===== parse|stringify ====
data/canada.json         510 MB/s   430 MB/s   670 MB/s   330 MB/s
data/citm_catalog.json  1110 MB/s   500 MB/s  1520 MB/s   650 MB/s
data/twitter.json        900 MB/s   610 MB/s   970 MB/s   650 MB/s
data/log.json           1080 MB/s  1080 MB/s  1080 MB/s  1080 MB/s

===== simd_json_tape ===== parse|stringify ===== parse|stringify ====
data/canada.json         760 MB/s
data/citm_catalog.json  1880 MB/s
data/twitter.json       1640 MB/s
data/log.json           2170 MB/s

[Licenser]

@mstump do you want to give this branch a try and see if it solves your issue?

[Licenser]

(marking this as non draft to get codecov to post it's stuff) - not ready to be merged yet but input is more then welcome

[codecov]

Codecov Report

Merging #88 into master will increase coverage by 1.07%.
The diff coverage is 86.87%.

@@            Coverage Diff             @@
##           master      #88      +/-   ##
==========================================
+ Coverage   84.65%   85.73%   +1.07%     
==========================================
  Files          36       39       +3     
  Lines        4385     5284     +899     
==========================================
+ Hits         3712     4530     +818     
- Misses        673      754      +81

Flag	Coverage Δ
#avx2	79.93% <86.87%> (+2.26%)	⬆️
#avx2KnownKey	78.68% <83.68%> (+1.04%)	⬆️
#e	?
#sse	80.57% <86.87%> (?)
#sseKnownKey	79.36% <83.68%> (+0.92%)	⬆️

Impacted Files	Coverage Δ
tests/jsonchecker.rs	100% <ø> (ø)	⬆️
src/value/borrowed/cmp.rs	89.36% <ø> (ø)	⬆️
src/value.rs	98.19% <ø> (+8.63%)	⬆️
src/value/owned/cmp.rs	87.5% <ø> (ø)	⬆️
src/value/borrowed/serialize.rs	80.28% <0%> (-4.8%)	⬇️
src/serde/value/borrowed/se.rs	0% <0%> (ø)	⬆️
src/value/owned.rs	100% <100%> (+4.48%)	⬆️
src/macros.rs	94.11% <100%> (-5.89%)	⬇️
src/value/owned/from.rs	89.83% <100%> (+1.15%)	⬆️
src/value/borrowed/from.rs	92.72% <100%> (+0.57%)	⬆️
... and 23 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6292255...51a2835. Read the comment docs.

[Licenser]

😂 well no this definetly needs work!

[Licenser]

@sunnygleason this should be ready for review :)

[Licenser]

@sunnygleason could I bug you for a review so we can merge this?

[mstump]

I finally climbed out from under my pile of demos and I'm sorry to report that I'm still getting weird behavior with this branch.

Cargo snipit:

[dependencies.simd-json]
git = "https://github.com/simd-lite/simdjson-rs"
branch = "128-bit-numbers"
features = ["128bit"]

With the 128bit feature enabled I get the error message:

thread 'reporter_message::raw_message::tests::test_deserialize_elastic_search' panicked at 'JsonSimdError(Error { index: 0, character: '💩', error: Serde("invalid type: i128, expected any value") })', src/reporter_message/raw_message.rs:398:25

Where previously I got:

thread 'reporter_message::raw_message::tests::test_deserialize_elastic_search' panicked at 'JsonSimdError(Error { index: 191048, character: '\n', error: Overflow })', src/reporter_message/raw_message.rs:398:25

The enum that it's decoding is

#[serde(untagged)]
#[derive(Clone, Debug, PartialEq, Deserialize, Serialize)]
pub enum RawMessageMetricValue {
    Boolean(bool),
    Float(f64),
    Long(i128),
    Integer(i64),
    Metric(Vec<RawMessageMetricNode>),
    String(String),
}

The offending JSON is below, it cleanly deserializes with serde_json.

{
  "name": "max_unsafe_auto_id_timestamp",
  "value": -9223372036854776000
}

[Licenser]

That looks like I did something wrong in the serde interop, I’ll check tomorrow if I can work out what serde expects of us for 128 but support thanks for sharing the structure and data that go right into a test Cheers, Heinz

…

________________________________ From: Matt Stump <[email protected]> Sent: Monday, January 13, 2020 10:45:23 PM To: simd-lite/simdjson-rs <[email protected]> Cc: Heinz Gies <[email protected]>; Author <[email protected]> Subject: Re: [simd-lite/simdjson-rs] 128 bit numbers (#88) I finally climbed out from under my pile of demos and I'm sorry to report that I'm still getting weird behavior with this branch. Cargo snipit: [dependencies.simd-json] git = "https://github.com/simd-lite/simdjson-rs" branch = "128-bit-numbers" features = ["128bit"] With the 128bit feature enabled I get the error message: thread 'reporter_message::raw_message::tests::test_deserialize_elastic_search' panicked at 'JsonSimdError(Error { index: 0, character: '💩', error: Serde("invalid type: i128, expected any value") })', src/reporter_message/raw_message.rs:398:25 Where previously I got: thread 'reporter_message::raw_message::tests::test_deserialize_elastic_search' panicked at 'JsonSimdError(Error { index: 191048, character: '\n', error: Overflow })', src/reporter_message/raw_message.rs:398:25 The enum that it's decoding is #[serde(untagged)] #[derive(Clone, Debug, PartialEq, Deserialize, Serialize)] pub enum RawMessageMetricValue { Boolean(bool), Float(f64), Long(i128), Integer(i64), Metric(Vec<RawMessageMetricNode>), String(String), } The offending JSON is below, it cleanly deserializes with serde_json. { "name": "max_unsafe_auto_id_timestamp", "value": -9223372036854776000 } — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#88>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAA5CNJTFOPL7VXWG3MVTKDQ5TOHHANCNFSM4JVOLEHQ>.

[Licenser]

Turns out this is a bug in serde, it doesn't support untagged fields with 128 bit values :/

https://github.com/serde-rs/serde/blob/ff70409215b62fa8cd138ce47534c432352c85bb/serde/src/private/de.rs#L230

[sunnygleason]

@Licenser what should the next step for this PR be?

[Licenser]

I opened a ticket with serde: serde-rs/serde#1717

I'm honestly not 100% sure, merging it would already allow 128 bit values outside of enums and should work there once it is fixed in serde.

Since no code changes will be required I think we could still merge it, there is little reason to hold it back just because of a bug in serde.

other then that clippy is throwing spanners in the works :P only 177 new errors! yay

[Licenser]

@sunnygleason can I bug you for a review so we can merge this? The remaining issue is out of our control and it will work once serde allows 128bit integers in untagged enums. So nothing we can do about that, our code is correct.

[sunnygleason]

@Licenser sure! Were you able to run fuzz testing for the feature? Just curious...

[Licenser]

I haven't it has been broken for a bit, and it feels with #95 in the todo list it's not worth investing time into unbreak it. We'll scrap the current fuzz stuff anyway.