Skip to content

Latest commit

 

History

History
206 lines (143 loc) · 7.85 KB

README.md

File metadata and controls

206 lines (143 loc) · 7.85 KB
Raw

python-autodict

This project aims to provide a simple file storage backend accessable by a dictionary interface. The dictionary like object syncs its contents with a file on disk.
Supports Python's pickle and JSON format.

Python's standard JSON encoder was improved to support Enums, NamedTuples, Tuples, bytes and Pathlib and some other types.

Basic usage

Just use it like a dict object but add a file path as first argument in your object instantiation.

from autodict import Autodict, FileFormat


# Create instance and load contents based on a file
mystore = Autodict("/tmp/empty_dict.myconf")
# Now use "mystore" like a dictionary.
# Call mystore.save() anytime. Else it will save right before destruction.



# Providing file format and default values for first use.
myconfig = Autodict("/tmp/default_dict.myconf", file_format=FileFormat.json_pretty, name="Python", count=2)
print(myconfig["name"])  # Use values immediately
# result:
#    Python (on first run)
#    Cool (on next run)

myconfig["name"] = "Cool"  # Save for next run

Constructor

instance = Autodict(
    file: Union[PathLike, str, bytes] = None,
    file_format: FileFormat = None,
    **defaults
)

file:
Path to a file (str, bytes or Path from pathlib). May be omitted. Then provide file on load or save methods.

file_format:
Specify the format of the file on disk from FileFormat enum. If omitted, class default Autodict.default_file_format will be used:

  • pickle_human: Human readable pickle (v0)
  • pickle_binary: Binary pickle v4 (Python 3.4)
  • json: Compacted Json dump
  • json_pretty: Pretty Json with indents

defaults (kwargs):
You may provide keyword args as default values for this instance. Keys will be treaten as strings. Other key types are not supported here due to **kwargs nature.

Advanced usage

You may change class default settings in the dirty way for all instances like this:

from autodict import Autodict, FileFormat
Autodict.default_file_format = FileFormat.json
instance = Autodict("myfile")

Just don't!

Better way is making use of inheritance to create specialized classes of autodict:

from autodict import Autodict, FileFormat

class MyConf(Autodict):
    __slots__ = ()  # Save memory by avoiding an extra unused dict
    default_content = {"type": "configfile", "configfile_version": 2}  # Default for each instance.
    auto_load = False  # Explicit call to load() preferred.
    default_file_format = FileFormat.json_pretty  # If you prefer json.

myconf = MyConf("myfile.json")

Class defaults

Alternative Autodict behaviour can be specified by subclassing from Autodict.

  • default_file_format: Which file format to use for instances without file_format specified.

  • auto_load: Calls load() after instantiation if True.

  • auto_mkdir: Allowes to create missing parent directories.

  • parents_create_mode: Filesystem attributes mode for created directories by auto_mkdir.

  • save_on_del: Calls save(force=True) right before deconstruction of instance if True.

  • track_changes: Tracks changes on immutable values if True. save() won't save if no change has made.

  • include_defaults:

    • False: Use defaults only if persistent file not found on load.
    • True: Use defaults and update from persistent file if exists.

  • default_content: Dictionary which will be copied and used for instances. Keys may be specified as any immutable data type. Constructor kwargs just support strings as keys.

  • expand_home: File paths containing "~" are expanded to the home directory.

  • auto_cast: Default is False: New values for existing keys will be cast to the original data types. If some_Autodict has an entry {"numeric": 1}, assigning some_Autodict["numeric"] = "2" will cast the string into int. Effective on restoring original data types from json files. Works on top level items only. Nested dicts won't be checked or cast.

  • bool_false, bool_true: Values which will be interpreted as False or True on active auto_cast. Default: {False, 0, "0", "n", "no", "false", 0., None} and {True, 1, "1", "y", "yes", "true"}

  • bool_default: Bool result on unknown types on active auto_cast. Default result: False.

  • noahs_ark_modules: Dangerzone. Affects following situation:

    • save_on_del enabled
    • Python interpreter is shutting down
    • save called by the destructor
    • Save format set to pickle_binary
    • Some classes like pathlib are used in the Autodict and want to be pickled.

    This situation fails because Python has already emptied sys.modules and pickle can't lookup modules and classes anymore.

    Solution is filling sys.modules again before pickle dumping. So if you want to pickle pathlib, put it into noahs_ark_modules like this:

    import pathlib
from autodict import Autodict

class MyConf(Autodict):
  # These following options are already default!  
  # save_on_del = True
  # default_file_format = FileFormat.pickle_binary
  __slots__ = ()  # Save memory by avoiding an extra unused dict
  default_content = {"path": pathlib.Path("/tmp")}  # Using pathlib
  noahs_ark_modules = {"pathlib": pathlib}  # Assign depending modules for dump pickle

Instance attributes

  • load(): Reloads data from the file. Unsaved data is lost.

  • save(): Saves current content of Autodict to file.

  • path: Property containing the path to load/save contents to/from. Contains path which was set by init.

  • All other methods from dict.

Properties

  • changed: Autodict tries to keep track of changed data inside. Unsafe. See restrictions.

Restrictions

  • Pickle mode may not support all classes.
  • JSON mode supports less data types of Python. Values may change or get lost. See auto_cast and next chapter.
  • Use with simple and immutable objects as data. int, str, float, bool, tuple, bytes, complex, frozenset, Decimal etc. Mutables like list, set, bytearray, dict are supported but changed attribute will not work. Use save(force=True) if you save it manually.

Data type handling in serialisation

Data type Pickle JSON
Pathlib (Path, PurePath) Native support Implemented
Enum Native support¹ Implemented²
tuple Native support Improved³
NamedTuple Native support¹ Implemented²
date, time, datetime Native support Implemented
bytes, bytearray Native support Implemented
set, frozenset Native support Implemented³

¹ Relies on the module path of the class as hint. Module containing the class must exist and has been loaded before calling load(). Classes may be moved to other modules.

² Relies primarily on classes used in default_content with same key. save() will also keep module path as fallback hint. Classes may be moved between modules.

³ Tuples, set and frozenset are not supported by JSON directly. This implementation encodes them into a list by prepending "__tuple__", "__set__" or "__frozenset__".

TODOs

  • tests
  • nested autodicts?
  • nested dicts with path access?
  • file extenstions
  • subclass name matching

Changelog

2021-02-28 v1.4

  • Implemented better support for Enums, Tuples, NamedTuples, datetime, date, time, bytes, bytearray, set, frozenset and Pathlib for JSON

2021-02-17 v1.3

  • auto_cast improved for booleans
  • fixed auto_cast was not applied on load()
  • bool_false, bool_true, bool_default
  • slots (reduced memory consumption)
  • parents_create_mode

2020-12-03 v1.2

  • expand_home setting
  • auto_mkdir
  • auto_cast
  • noahs_ark_modules workaround for dangerous auto save of objects to pickle on interpreter shutdown

2020-11-16 v1.0

  • Implemented json, pickle, class settings
  • Added class level default values
  • Added interchangable file
  • Single r+w file handle which is keeped open until instance destruction
  • Simplified track_changes
  • First manual tests
  • Added licence