Skip to content

Files

Latest commit

 

History

History
137 lines (92 loc) · 3.73 KB

README.md

File metadata and controls

137 lines (92 loc) · 3.73 KB

📖 A Python json-filelogger

Lint code and run tests pypi

A simple way to log data to json-files from within your code.

Installation

You can use the python package manager (pip) to install the file-logger:

pip install jsonfilelogger

Usage: LogWriter

Create a LogWriter:

from jsonfilelogger.logger import LogWriter

writer = LogWriter(folder=".", filename="log.json", threshold=10)

Parameters

folder: the folder where the logfile should be stored

filename: the name of the logfile

threshold: the threshold for auto-flushing; if set None the writer will not autoflush

Example

Log data to the logfile:

writer.log({"key_int": 1,
            "key_string": "logger",
            "key_boolean": True,
            "key_double": 1.337,
            "key_list": ["element1", "element2"],
            "key_none": None})

Usage: LogReader

Create a LogReader:

from jsonfilelogger.logger import LogReader

reader = LogReader(folder="./", filename="log.json")

Parameters

folder: the folder where the logfile should be stored

filename: the name of the logfile

Example

Read data from the logfile:

data = reader.retrieve()
... # do something with the data

Methods

Class Method Explanation
LogReader, LogWriter .reset() This method clears the contents of the logfile.
LogReader, LogWriter .remove() This methode removes the logfile.
LogWriter .with_datetime(folder: str, threshold: int) This method creates a logfile in the folder with the current timestamp as filename.
LogWriter .log(data) This method logs the data (dictionary) in the logfile.
LogWriter .flush() This method flushes the buffer in the writer. The threshold of the buffer is given in the constructor
LogReader .retrieve() This method retrieves all the data from the logfile and returns it as a list of dictionaries.

Demo usage

An example of how to use this logger is given. Imagine one has created a major breakthrough AI system that still has to be trained. During training one want's to keep an eye on the performance of the progress. To do this, the LogWriter can be added in the learning iterations (e.g. in one Jupyter notebook). In another process (e.g. in another Jupyter notebook) it is then possible to read in the data and make amazing visualisations (yay, visualisations!) of how your breakthrough model is performing...

from jsonfilelogger.logger import LogWriter

writer = LogWriter.with_datetime(folder="./", threshold=10)

super_great_ai_model = ...
iteration = 0
while iteration < 100_000:
    score, other_data = super_great_ai_model.do_iteration()
    writer.log({
        score: score,
        other_data: other_data
    })
    iteration += 1
writer.flush()
from jsonfilelogger.logger import LogReader
import glob

logs_files = sorted(glob.glob("./[0-9]*.json"))
reader = LogReader(folder="./", filename=logs_files[-1])

data = reader.retrieve()

# Make awesome visualisations!
...

There may also be other use cases, you can use this logger as you want!

Tests

You can run the tests yourselves by:

cd tests
pip3 install -r requirements.txt
pytest

License

This software is licensed under the MIT license.

Bugs/ Ideas?

In case of questions, ideas or bugs you can always create an issue or contact me.