jing 0.3.3

Download stock data and perform data analisys

jing

This is a python library, used for download stock data and perform data analysis.

Installation

pip install jing

All required dependencies (requests, pandas, baostock, yfinance, akshare, pyarrow) will be installed automatically.

Usage

Only 2 classes are exported.

downloader D

• Function: download the data from internet to local file system
• Parameter:
  • _market (str): us, cn, hk, default is us
• Return Value: No return value

sample code

import jing

# US stock via Yahoo Finance
d = jing.D('us')
d.download('AAPL')

# CN stock via BaoStock (default source for CN)
d = jing.D('cn')
d.download('sz.000807')

# HK stock via AkShare
d = jing.D('hk')
d.download('00700')

Data layout

The project now uses a source-first layout under data/:

data/
  raw/
    yahoo/us/
    baostock/cn/
    akshare/cn/
    akshare/hk/
    binance/bn/
    list/
  parquet/
  duckdb/

Examples:

• US Yahoo CSV: data/raw/yahoo/us/AAPL.csv
• CN BaoStock CSV: data/raw/baostock/cn/sh.601088.csv
• HK AkShare CSV: data/raw/akshare/hk/00700.csv
• list files: data/raw/list/us.txt

Configuration

Data directory

Priority (highest first):

1. Function call — programmatic override

   from jing.data_paths import set_data_root
   set_data_root('/custom/data/path')
   

2. Environment variable

   export JING_DATA=/custom/data/path
   

3. Default — ~/data/jing

Cache

jing caches fetched stock data as Parquet files to speed up repeated analysis.

• Cache dir: ~/.cache/jing/data (override with JING_CACHE env var)
• Cache key: market_source_code_date.parquet

The date in the cache key is the analysis date you pass (e.g. _date='2024-01-01'). If no date is passed, today's date is used. This ensures cached data is scoped to the specific date filter you requested.

Why Parquet?

• Fast reads — columnar format, much faster than CSV for large DataFrames
• Small size — binary + compression (~45% smaller than CSV in our tests)
• Preserves schema — dates, floats, and dtypes survive round-trip without re-parsing
• Pandas-native — pd.read_parquet() / df.to_parquet() with no extra code

Trade-off: requires pyarrow dependency. For repeated stock data reads the speedup is worth it.

Skip cache

y = jing.Y('sh.600519', _date='2024-01-01', _market='cn', skip_cache=True)

Clear cache

from jing.data_center import DataCenter

dc = DataCenter('cn', 'baostock')
dc.clear_cache()                               # clear all
dc.clear_cache(code='sh.600519')               # clear specific stock
dc.clear_cache(code='sh.600519', date='2024-01-01')  # clear specific entry

Release to PyPI

1. Bump the version in pyproject.toml:

   version = "0.3.1"
   

2. Build the distribution packages:

   python -m build
   

3. Upload to PyPI:

   python -m twine upload dist/*
   

   You will need a PyPI account and API token. Configure twine once with python -m twine upload --repository testpypi dist/* to test first if desired.