Skip to content

A plug-and-play JIT implementation for Marshmallow to speed up data serialization and deserialization

License

Notifications You must be signed in to change notification settings

mLupine/DeepFriedMarshmallow

Β 
Β 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

32 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

πŸ”₯ Deep-Fried Marshmallow – Makes Marshmallow a Chicken Nugget

I need to be honest with you β€” I have no idea how to compare the speed of a marshmallow and the speed of a chicken nugget. I really liked that headline, though, so let's just assume that a nugget is indeed faster than a marshmallow. So is this project, Deep-Fried Marshmallow, faster than vanilla Marshmallow. Or, to be precise, it makes Marshmallow faster.

Deep-Fried Marshmallow implements a JIT for Marshmallow that speeds up dumping objects 3-5x (depending on your schema). Deep-Fried Marshmallow allows you to have the great API that Marshmallow provides without having to sacrifice performance.

    Benchmark Result:
        Original Dump Time: 220.50 usec/dump
        Original Load Time: 536.51 usec/load
        Optimized Dump Time: 58.67 usec/dump
        Optimized Load Time: 118.44 usec/load

        Speed up for dump: 3.76x
        Speed up for load: 4.53x

Deep-Fried Marshmallow is a fork of the great Toasted Marshmallow project that, sadly, has been abandoned for years. Deep-Fried Marshmallow introduces many changes that make it compatible with all latest versions of Marshmallow (3.13+). It also changes the way the library interacts with Marshmallow, which means that code of Marshmallow doesn't need to be forked and modified for the JIT magic to work. That's a whole new level of magic!

Installing Deep-Fried Marshmallow

pip install DeepFriedMarshmallow
# or, if your project uses Poetry:
poetry install DeepFriedMarshmallow

If your project doesn't have vanilla Marshmallow specified in requirements, the latest version of it will be installed alongside Deep-Fried Marshmallow. You are free to pin any version of it that you need, as long as it's newer than v3.13.

Enabling Deep-Fried Marshmallow

Enabling Deep-Fried Marshmallow on an existing schema is just one change of code. Change your schemas to inherit from the JitSchema class in the deepfriedmarshmallow package instead of Schema from marshmallow.

For example, this block of code:

from marshmallow import Schema, fields

class ArtistSchema(Schema):
    name = fields.Str()

class AlbumSchema(Schema):
    title = fields.Str()
    release_date = fields.Date()
    artist = fields.Nested(ArtistSchema())

schema = AlbumSchema()

Should become this:

from marshmallow import fields
from deepfriedmarshmallow import JitSchema

class ArtistSchema(JitSchema):
    name = fields.Str()

class AlbumSchema(JitSchema):
    title = fields.Str()
    release_date = fields.Date()
    artist = fields.Nested(ArtistSchema())

schema = AlbumSchema()

And that's it!

Auto-patching all Marshmallow schemas

If you want to automatically patch all Marshmallow schemas in your project, Deep-Fried Marshmallow provides a helper function for that. Just call deepfriedmarshmallow.deep_fry_marshmallow() before you start using Marshmallow schemas, and you're all set. The upmost __init__.py file of your project is a good place to do that.

# your_package/__init__.py
from deepfriedmarshmallow import deep_fry_marshmallow

deep_fry_marshmallow()

All imports of marshmallow.Schema will be automatically replaced with deepfriedmarshmallow.Schema with no other changes to your code. Isn't that sweet extra crispy?

Custom Schema classes

Deep-Fried Marshmallow also provides a mixin class that you can use to create or extend custom Schema classes. To use it, just inherit from JitSchemaMixin. Let's take a look at the following example:

from marshmallow import fields

class ClockSchema(MyAwesomeBaseSchema):
    time = fields.DateTime(data_key="Time")

If you want to make this schema JIT-compatible, and don't want to modify the MyAwesomeBaseSchema class to inherit from deepfriedmarshmallow.Schema, you can do the following:

from marshmallow import fields
from deepfriedmarshmallow import JitSchemaMixin

class ClockSchema(JitSchemaMixin, MyAwesomeBaseSchema):
    time = fields.DateTime(data_key="Time")

Patcher functions

If all of the above wasn't enough, Deep-Fried Marshmallow also provides two more ways to patch Marshmallow schemas. Both of them are functions that you can call to patch either a Schema class, or a Schema instance. Let's take a look at the following example:

from marshmallow import Schema, fields
from deepfriedmarshmallow import deep_fry_schema

class ArtistSchema(Schema):
    name = fields.Str()

deep_fry_schema(ArtistSchema)
schema = ArtistSchema()

The deep_fry_schema function will patch the AlbumSchema class, and all instances of it will be JIT-compatible. If you want to patch a specific instance of a schema, you can use the deep_fry_schema_object function:

from marshmallow import Schema, fields
from deepfriedmarshmallow import deep_fry_schema_object

class ArtistSchema(Schema):
    name = fields.Str()

schema = ArtistSchema()
deep_fry_schema_object(schema)

This function will patch the schema object, and all dumps and loads will be JIT-compatible. This function is useful if you want to patch a schema that you don't have control over, for example, a schema that is provided by a third-party library.

How it works

Deep-Fried Marshmallow works by generating code at runtime to optimize dumping objects without going through layers and layers of reflection. The generated code optimistically assumes the objects being passed in are schematically valid, falling back to the original Marshmallow code on failure.

For example, taking AlbumSchema from above, Deep-Fried Marshmallow will generate the following methods:

def InstanceSerializer(obj):
    res = {}
    value = obj.title; value = value() if callable(value) else value; value = str(value) if value is not None else None; res["title"] = value
    value = obj.release_date; value = value() if callable(value) else value; res["release_date"] = _field_release_date__serialize(value, "release_date", obj)
    value = obj.artist; value = value() if callable(value) else value; res["artist"] = _field_artist__serialize(value, "artist", obj)
    return res

def DictSerializer(obj):
    res = {}
    if "title" in obj:
        value = obj["title"]; value = value() if callable(value) else value; value = str(value) if value is not None else None; res["title"] = value
    if "release_date" in obj:
        value = obj["release_date"]; value = value() if callable(value) else value; res["release_date"] = _field_release_date__serialize(value, "release_date", obj)
    if "artist" in obj:
        value = obj["artist"]; value = value() if callable(value) else value; res["artist"] = _field_artist__serialize(value, "artist", obj)
    return res

def HybridSerializer(obj):
    res = {}
    try:
        value = obj["title"]
    except (KeyError, AttributeError, IndexError, TypeError):
        value = obj.title
    value = value; value = value() if callable(value) else value; value = str(value) if value is not None else None; res["title"] = value
    try:
        value = obj["release_date"]
    except (KeyError, AttributeError, IndexError, TypeError):
        value = obj.release_date
    value = value; value = value() if callable(value) else value; res["release_date"] = _field_release_date__serialize(value, "release_date", obj)
    try:
        value = obj["artist"]
    except (KeyError, AttributeError, IndexError, TypeError):
        value = obj.artist
    value = value; value = value() if callable(value) else value; res["artist"] = _field_artist__serialize(value, "artist", obj)
    return res

Deep-Fried Marshmallow will invoke the proper serializer based upon the input.

Since Deep-Fried Marshmallow generates code at runtime, it's critical you re-use Schema objects. If you're creating a new Schema object every time you serialize or deserialize an object, you're likely to experience much worse performance.

Special thanks to

  • @rowillia/@lyft β€” for creating Toasted Marshmallow
  • @taion β€” for a PoC of injecting the JIT compiler by replacing the marshaller
  • @Kalepa β€” for needing improved Marshmallow performance so that I could actually work on this project πŸ˜…

License

See LICENSE for details.

Contributing

Contributions, issues and feature requests are welcome!

Feel free to check existing issues before reporting a new one.

Show your support

Give this repository a ⭐️ if this project helped you!

About

A plug-and-play JIT implementation for Marshmallow to speed up data serialization and deserialization

Topics

Resources

License

Stars

Watchers

Forks

Languages

  • Python 100.0%