/hermes

A Haskell library for fast, memory-efficient decoding of JSON documents using the simdjson C++ library

Primary LanguageC++MIT LicenseMIT

hermes

CI badge Hackage badge

A Haskell interface over the simdjson C++ library for decoding JSON documents. Hermes, messenger of the gods, was the maternal great-grandfather of Jason, son of Aeson.

Overview

This library exposes functions that can be used to write decoders for JSON documents using the simdjson On Demand API. From the simdjson On Demand design documentation:

Good applications for the On Demand API might be:

You are working from pre-existing large JSON files that have been vetted. You expect them to be well formed according to a known JSON dialect and to have a consistent layout. For example, you might be doing biomedical research or machine learning on top of static data dumps in JSON.

Both the generation and the consumption of JSON data is within your system. Your team controls both the software that produces the JSON and the software the parses it, your team knows and control the hardware. Thus you can fully test your system.

You are working with stable JSON APIs which have a consistent layout and JSON dialect.

With this in mind, Data.Hermes parsers can potentially decode Haskell types faster than traditional Data.Aeson.FromJSON instances, especially in cases where you only need to decode a subset of the document. This is because Data.Aeson.FromJSON converts the entire document into a Data.Aeson.Value, which means memory usage increases linearly with the input size. The simdjson::ondemand API does not have this constraint because it iterates over the JSON string in memory without constructing an intermediate tree. This means decoders are truly lazy and you only pay for what you use.

Usage

This library does not offer a Haskell API over the entire simdjson On Demand API. It currently binds only to what is needed for defining and running a Decoder. You can see the tests and benchmarks for example usage. Decoder a is a thin layer over IO that keeps some context around for better error messages. simdjson::ondemand exceptions will be caught and re-thrown with enough information to troubleshoot. In the worst case you may run into a segmentation fault that is not caught, which you are encouraged to report as a bug.

Decoders

personDecoder :: Value -> Decoder Person
personDecoder = withObject $ \obj ->
  Person
    <$> atKey "_id" text obj
    <*> atKey "index" int obj
    <*> atKey "guid" text obj
    <*> atKey "isActive" bool obj
    <*> atKey "balance" text obj
    <*> atKey "picture" (nullable text) obj
    <*> atKey "latitude" scientific obj

-- Decode a strict ByteString.
decodePersons :: ByteString -> Either HermesException [Person]
decodePersons = decodeEither $ list personDecoder

It looks a little like Waargonaut.Decode.Decoder m, just not as polymorphic. The interface is copied because it's elegant and does not rely on typeclasses. However, hermes does not give you a cursor to play with, the cursor is implied and is forward-only (except when accessing object fields). This limitation allows us to write very fast decoders.

Exceptions

When decoding fails for a known reason, you will get a Left HermesException indicating if the error came from simdjson or from an internal hermes call. The exception contains a DocumentError record with some useful information, for example:

*Main> decodeEither (withObject . atKey "hello" $ list text) "{ \"hello\": [\"world\", false] }" 
Left (SIMDException (DocumentError {path = "/hello/1", errorMsg = "Error while getting value of type text. The JSON element does not have the requested type.", docLocation = "false] }", docDebug = "json_iterator [ depth : 3, structural : 'f', offset : 21', error : No error ]"}))

Benchmarks

We benchmark the following operations using both hermes-json and aeson strict ByteString decoders:

  • Decode an array of 1 million 3-element arrays of doubles
  • Decode a very small object into a Map
  • Full decoding of a large-ish (12 MB) JSON array of objects
  • Partial decoding of Twitter status objects to highlight the on-demand benefits

Intel Core i7-7500U @2.70GHz / 2x8GB RAM @LPDDR3

Non-threaded runtime

Name Mean (ps) 2*Stdev (ps) Allocated Copied Peak Memory
All.1 Million 3-Arrays.Hermes [[Double]] 515111634200 19332030346 567061009 555768600 548405248
All.1 Million 3-Arrays.Aeson [[Double]] 1879357624000 109849580814 9240070406 918469928 815792128
All.Small Object to Map.Hermes Decode 1453784 99664 4509 143 815792128
All.Small Object to Map.Aeson Lazy 2936816 184144 20403 1 815792128
All.Small Object to Map.Aeson Strict 2954601 167788 20379 1 815792128
All.Full Persons Array.Ordered Keys.Hermes Decode 106374445200 1634370706 150384445 77972337 815792128
All.Full Persons Array.Ordered Keys.Aeson Lazy 449523758200 41027868974 1212515956 268045285 815792128
All.Full Persons Array.Ordered Keys.Aeson Strict 353756693800 33058983332 1212278538 200617993 815792128
All.Full Persons Array.Unordered Keys.Hermes Decode 113706475400 3830951512 150322956 75438334 815792128
All.Full Persons Array.Unordered Keys.Aeson Lazy 439971868800 7835893708 1213065712 264393596 815792128
All.Full Persons Array.Unordered Keys.Aeson Strict 353582487600 34203690928 1212032626 200502732 815792128
All.Partial Twitter.Hermes Decode 448839050 22309802 398892 4643 815792128
All.Partial Twitter.Aeson Lazy 16644838050 1080972950 52866100 7541778 815792128
All.Partial Twitter.Aeson Strict 14025762400 793325836 53294115 5958334 815792128

Threaded runtime

Name Mean (ps) 2*Stdev (ps) Allocated Copied Peak Memory
All.1 Million 3-Arrays.Hermes [[Double]] 541886019300 28985190366 567061372 555825639 547356672
All.1 Million 3-Arrays.Aeson [[Double]] 1948081269800 127560261526 9240069829 919131304 815792128
All.Small Object to Map.Hermes Decode 1443894 82886 4262 144 815792128
All.Small Object to Map.Aeson Lazy 3010568 205200 20404 2 815792128
All.Small Object to Map.Aeson Strict 3012037 99100 20405 2 815792128
All.Full Persons Array.Ordered Keys.Hermes Decode 112708023300 5172458580 150387033 79494634 815792128
All.Full Persons Array.Ordered Keys.Aeson Lazy 460293246300 2987469998 1212937927 264367279 815792128
All.Full Persons Array.Ordered Keys.Aeson Strict 360104741850 6231191600 1212295571 192141024 815792128
All.Full Persons Array.Unordered Keys.Hermes Decode 118564635400 8355435606 150325103 76318688 815792128
All.Full Persons Array.Unordered Keys.Aeson Lazy 476108313200 33933903066 1212647160 268222585 815792128
All.Full Persons Array.Unordered Keys.Aeson Strict 363291220000 15069371922 1212036871 190951202 815792128
All.Partial Twitter.Hermes Decode 465476487 6501684 402833 4861 815792128
All.Partial Twitter.Aeson Lazy 17562410575 370107430 53230448 7657222 815792128
All.Partial Twitter.Aeson Strict 14433454400 1426192330 53290220 6050622 815792128

Performance Tips

  • Decode to Text instead of String wherever possible!
  • Decode to Int or Double instead of Scientific if you can.
  • Decode your object fields in order. Out of order field lookups will slightly degrade performance. If encoding with aeson, you can leverage toEncoding to enforce ordering.
  • You can improve performance by holding onto your own HermesEnv. decodeEither creates and destroys the simdjson instances every time it runs, which adds a performance penalty. Beware, do not share a HermesEnv across multiple threads.

Limitations

Because the On Demand API uses a forward-only iterator (except for object fields), you must be mindful to not access values out of order. In other words, you should not hold onto a Value to parse later since the iterator may have already moved beyond it.

Because the On Demand API does not validate the entire document upon creating the iterator (besides UTF-8 validation and basic well-formed checks), it is possible to parse an invalid JSON document but not realize it until later. If you need the entire document to be validated up front then a DOM parser is a better fit for you.

The On Demand approach is less safe than DOM: we only validate the components of the JSON document that are used and it is possible to begin ingesting an invalid document only to find out later that the document is invalid. Are you fine ingesting a large JSON document that starts with well formed JSON but ends with invalid JSON content?

This library currently cannot decode scalar documents, e.g. a single string, number, boolean, or null as a JSON document.

Portability

Per the simdjson documentation:

A recent compiler (LLVM clang6 or better, GNU GCC 7.4 or better, Xcode 11 or better) on a 64-bit (PPC, ARM or x64 Intel/AMD) POSIX systems such as macOS, freeBSD or Linux. We require that the compiler supports the C++11 standard or better.

However, this library relies on std::string_view without a shim, so C++17 or better is highly recommended.