Jacksung

jacksung is a personal Python utility package for research and engineering workflows. It includes helpers for ECNU network login, logging, multithreading, MySQL access, NumPy/GeoTIFF/NetCDF conversion, image processing, NVIDIA GPU monitoring, LaTeX polishing with LLMs, and several AI/meteorological data utilities.

Python 3.9 or later is required. Python 3.11 is recommended.

Installation

Create a clean conda environment first if needed:

conda create -n jacksung python=3.11
conda activate jacksung
conda install gdal==3.6.2
pip install jacksung

For local development:

pip install -r requirements.txt

Package Layout

jacksung/
  utils/        General utilities: login, log, database, time, image, conversion, GPU monitor
  ai/           AI helpers: metrics, LaTeX polishing, GeoNet/GeoAttX, satellite data utilities

Main command line tools:

• ecnu_login: log in, check, or log out of the ECNU campus network.
• watch_gpu: print NVIDIA GPU status using nvidia-smi.

Feature Index

The package contains the following modules and utilities.

General utilities:

• jacksung.utils.base_db: MySQL connection wrapper, SQL execution, string/number conversion helpers.
• jacksung.utils.cache: small in-memory FIFO cache with keyed values.
• jacksung.utils.data_convert: NetCDF, NumPy, TIFF/GeoTIFF conversion, coordinate helpers, DMS conversion, lon/lat transform fitting, NaN window filling.
• jacksung.utils.exception: custom file/NaN exceptions and wait_fun retry helper.
• jacksung.utils.fastnumpy: fast NumPy save/load plus binary pack/unpack helpers and streaming mean accumulator.
• jacksung.utils.figure: figure rendering helpers for NumPy grids, color bars, labeled maps, and image export.
• jacksung.utils.hash: file, file-list, and string hashing.
• jacksung.utils.image: pixel lookup by coordinate, text drawing, borders, color maps, crop, concat, GIF, zoom/dock, and boundary extraction.
• jacksung.utils.log: timestamped print, server log sender, stdout tee/file logger.
• jacksung.utils.login: Selenium-based ECNU login client and ecnu_login CLI.
• jacksung.utils.mean_std: merge partial mean/std statistics and compute mean/std from accumulated sums.
• jacksung.utils.multi_task: thread/process task executor and lock helpers.
• jacksung.utils.number: numeric formatting helpers.
• jacksung.utils.nvidia: colored nvidia-smi display and watch_gpu CLI.
• jacksung.utils.time: date/time string helpers, remaining-time estimator, stopwatch, human-readable size formatting.
• jacksung.utils.web: Chrome Selenium driver factory with headless, temporary directory, and download directory options.

AI and meteorological utilities:

• jacksung.ai.latex_tool: OpenAI-compatible LaTeX polishing workflow, prompt builders, merge/diff helpers.
• jacksung.ai.metrics: precipitation metrics, bootstrap uncertainty, RMSE, PSNR, SSIM, AUROC, tensor conversion.
• jacksung.ai.GeoAttX: GeoAttX base class and prediction workflows for interpolation, precipitation/QPE, and Huayu-style inference.
• jacksung.ai.GeoNet: GeoNet network definitions and reusable model blocks.
• jacksung.ai.utils.cmorph: CMORPH HDF to NumPy conversion.
• jacksung.ai.utils.data_parallelV2: balanced PyTorch DataParallel helpers.
• jacksung.ai.utils.fy: FY satellite coordinate tools, filename parsing, HDF/NetCDF conversion, clipping, registration, and date lookup.
• jacksung.ai.utils.fy3g: FY-3G HDF conversion and filename parsing.
• jacksung.ai.utils.goes: GOES resampling, single-channel extraction, directory lookup, and NumPy conversion.
• jacksung.ai.utils.gsmap: GSMaP HDF to NumPy conversion.
• jacksung.ai.utils.imerg: IMERG downloader and HDF to NumPy conversion.
• jacksung.ai.utils.metsat: Meteosat/SEVIRI NAT processing through Satpy, WGS84 area definition, and file lookup.
• jacksung.ai.utils.norm_util: prediction, precipitation, and generic normalization helpers.
• jacksung.ai.utils.util: model loading/saving, config parsing, device transfer, metric tracking, plotting, augmentation, and satellite clipping.

ECNU Login

The login tool uses Selenium and ChromeDriver. Install Chrome first, then download the matching ChromeDriver from Chrome for Testing.

Expected driver layout:

Home directory
`-- chrome
    |-- chromedriver.exe   # Windows
    |-- chromedriver       # Linux/macOS
    `-- tmp

On Windows the default path is ~/chrome/chromedriver.exe; on other systems it is ~/chrome/chromedriver.

Run with credentials:

ecnu_login -u account -p password

Or store credentials in ~/.ecnu_login:

u: account
p: password

Then run:

ecnu_login

Supported actions:

ecnu_login -t login_check
ecnu_login -t login
ecnu_login -t logout

Python usage:

from jacksung.utils.login import ecnu_login

login = ecnu_login(driver_path="chromedriver_path", tmp_path="tmp_path")
login.get_drive()
login.login_check("username", "password")
login.login("username", "password")
login.logout()
login.close_driver()

Logging

Print messages with timestamps:

from jacksung.utils.log import oprint as print

print("this is a log")

Send log messages to a server. The URL should accept name and content parameters, for example https://www.example.com?log&api-key=123&name=logname&content=logcontent.

from jacksung.utils.log import LogClass

log_class = LogClass(on=True, url="https://www.example.com?log&api-key=123")
log_class.send_log("35.8", "PSNR")

Record terminal output to files:

import sys
from jacksung.utils.log import StdLog

if __name__ == "__main__":
    sys.stdout = StdLog(filename="log.txt", common_path="warning.txt")
    print("[TemporaryTag]Only in terminal", end="[TemporaryTag]\n")
    print("[Warning]In warning.txt and terminal", end="[Warning]\n")
    print("[Error]In warning.txt and terminal", end="[Error]\n")
    print("[Common]Common in warning.txt and terminal", end="[Common]\n")
    print("[OnlyFile]OnlyFile in warning.txt and terminal", end="[OnlyFile]\n")
    print("In log.txt and terminal")

Multithreading

import time
from jacksung.utils.multi_task import MultiTasks

def worker(idx):
    print(idx)
    time.sleep(2)
    return idx

mt = MultiTasks(threads=3)
for idx in range(10):
    mt.add_task(idx, worker, [idx])

results = mt.execute_task()

MultiTasks uses a thread pool by default. Process mode is also available through pool=type_process from jacksung.utils.multi_task.

Fast NumPy

import jacksung.utils.fastnumpy as fnp

data = fnp.load("data.npy")
fnp.save("copy.npy", data)

MySQL

BaseDB reads connection settings from an ini file and executes SQL through PyMySQL.

from jacksung.utils.base_db import BaseDB

class DB:
    def __init__(self, ini_path="db.ini"):
        self.bd = BaseDB(ini_path)

    def insert_record(self, year, month, day):
        sql = f"INSERT INTO `data_record` (`year`, `month`, `day`) VALUES ({year}, {month}, {day});"
        self.bd.execute(sql)

    def select_record(self, year, month, day):
        sql = f"SELECT COUNT(1) FROM data_record WHERE year={year} AND month={month} AND day={day};"
        result, cursor = self.bd.execute(sql)
        return cursor.fetchone()[0]

Example db.ini:

[database]
host = 127.0.0.1
user = root
password = root
database = XXXX

NVIDIA GPU Monitor

Print GPU information:

watch_gpu

On Linux, you can wrap it with watch:

alias watch-gpu="watch -n 1 -d watch_gpu"

Time Utilities

Estimate remaining time:

import time
from jacksung.utils.time import RemainTime

epochs = 100
rt = RemainTime(epochs)

for _ in range(epochs):
    rt.update()
    time.sleep(2)

Use a stopwatch:

import time
from jacksung.utils.time import Stopwatch

sw = Stopwatch()
time.sleep(1)
print(sw.pinch())  # elapsed time since last reset
time.sleep(1)
print(sw.reset())  # elapsed time and reset

Data Conversion

Convert NetCDF to NumPy and NumPy arrays to TIFF/GeoTIFF.

import numpy as np
from jacksung.utils.data_convert import nc2np, np2tif

nc_data, dim = nc2np(r"C:\Users\ECNU\Desktop\upper.nc")

# Without geocoordinates
np2tif(nc_data, "constant_masks/upper", dim_value=dim)

# With geocoordinates
np2tif(
    "constant_masks/land_mask.npy",
    save_path="constant_masks",
    out_name="land_mask",
    left=0,
    top=90,
    x_res=0.25,
    y_res=0.25,
    dtype=np.float32,
)

dim_value can be used to name output files generated from multi-dimensional arrays:

dim_value = [{"value": ["WIN", "TMP"]}, {"value": ["PRS", "HEIGHT"]}]

Other useful helpers include Coordinate, nc2tif, get_transform_from_lonlat_matrices, haversine_distance, and fill_nan_with_window_mean.

Figure Utilities

jacksung.utils.figure focuses on converting NumPy arrays into visual products with color mapping and optional geospatial context.

from jacksung.utils.figure import make_color_map, make_fig

colors = [
    [0, "#FFFFFF"],
    [10, "#00A0FF"],
    [50, "#FFDD00"],
    [100, "#FF0000"],
]

color_bar = make_color_map(colors, h=220, w=1200, unit="mm")
make_fig(
    "rain.npy",
    area=((100, 140, 10), (20, 60, 10)),
    save_name="figures/rain.png",
    colors=colors,
    colormap_unit="mm",
)

Image Utilities

jacksung.utils.image includes helpers for drawing text and borders, building color maps, cropping PNGs, concatenating images, creating GIFs, zooming image regions, and drawing boundaries.

import cv2
from jacksung.utils.image import concatenate_images, create_gif

img1 = cv2.imread("a.png")
img2 = cv2.imread("b.png")
merged = concatenate_images([img1, img2], direction="h")
cv2.imwrite("merged.png", merged)

create_gif("frames_dir", "demo.gif", duration=500)

Web Driver

Create a Chrome Selenium driver for browser automation or downloads:

from jacksung.utils.web import make_driver

driver = make_driver(
    url="https://example.com",
    is_headless=True,
    tmp_path="chrome_tmp",
    download_dir="downloads",
)
driver.quit()

Cache, Retry, and Statistics Helpers

import numpy as np
from jacksung.utils.cache import Cache
from jacksung.utils.exception import wait_fun
from jacksung.utils.mean_std import cal_mean_std_one_loop, mean_std_part2all

cache = Cache(cache_len=2)
cache.add_key("a", 1)
print(cache.get_key_in_cache("a"))

result = wait_fun(lambda x: x + 1, args=[1])

batch = np.random.rand(4, 3, 16, 16)
s = batch.sum(axis=0)
ss = (batch ** 2).sum(axis=0)
mean_pixel, std_pixel, mean_level, std_level = cal_mean_std_one_loop(s, ss, count=4)
merged_mean, merged_std = mean_std_part2all([4], [mean_pixel], [std_pixel])

AI Tools

LaTeX Auto Polish

jacksung.ai.latex_tool.polish can polish a LaTeX manuscript through an OpenAI-compatible LLM server.

from jacksung.ai.latex_tool import polish

polish(
    main_dir_path="your latex root directory",
    tex_file="main.tex",
    server_url="The full LLM server url with /v1",
    token="Your token here",
)

Notes:

• If the paper is Chinese or needs a Chinese prompt, set cn_prompt=True.
• To use a custom prompt, pass prompt containing {text}.
• The tool generates old.tex, new.tex, and diff.tex in the parent directory.
• The change-tracking PDF is compiled from diff.tex.
• If diff.tex fails to compile, fix the generated new.tex first.
• A strong model is recommended; small models may introduce LaTeX syntax errors.

Metrics and Meteorological Utilities

The jacksung.ai package also contains:

• jacksung.ai.metrics: precipitation metrics, RMSE, PSNR, SSIM, AUROC, and bootstrap uncertainty.
• jacksung.ai.GeoNet: GeoNet model definitions.
• jacksung.ai.GeoAttX: prediction helpers for GeoAttX-related workflows.
• jacksung.ai.utils: satellite and precipitation data utilities for FY, FY-3G, GOES, GSMaP, IMERG, CMORPH, SEVIRI/Meteosat, normalization, and PyTorch training helpers.

These modules depend on scientific and geospatial packages such as NumPy, rasterio, netCDF4, satpy, pyresample, OpenCV, Pillow, and PyTorch.

Training Utilities

import numpy as np
from jacksung.ai.utils.util import data_to_device, load_model, parse_config, save_model
from jacksung.ai.utils.norm_util import Normalization

config = parse_config("config.yml")
mean_std = np.load("mean_std.npy")
norm = Normalization(mean_std)

BalancedDataParallel can be used when GPU 0 should receive a smaller batch than other GPUs.

from jacksung.ai.utils.data_parallelV2 import BalancedDataParallel

model = BalancedDataParallel(gpu0_bsz=2, module=model, device_ids=[0, 1, 2])

Satellite Data Utilities

The satellite helpers convert common precipitation and meteorological products into NumPy arrays and provide filename/date/coordinate utilities.

from datetime import datetime
from jacksung.ai.utils.fy import getNPfromHDF as read_fy
from jacksung.ai.utils.goes import getNPfromDir as read_goes_dir
from jacksung.ai.utils.metsat import getNPfromNAT

date = datetime(2024, 1, 1, 0, 0)
fy_data = read_fy("FY4A_file.HDF")
goes_data = read_goes_dir("goes_dir", date)
metsat_data = getNPfromNAT("seviri_file.nat")

Hash and Miscellaneous Utilities

from jacksung.utils.hash import calculate_file_hash, hash_string
from jacksung.utils.number import round2str

print(hash_string("hello"))
print(calculate_file_hash("README.md"))
print(round2str(3.14159, digits=2))

More hash helpers:

from jacksung.utils.hash import hash_files

digest = hash_files(["README.md", "setup.py"])

Development and Release （for developers）

Build and upload a release:

python setup.py sdist bdist_wheel
twine upload dist/*

Be aware that setup.py currently increments the local version stored in loacaldb.json, removes build artifacts, and attempts to commit the version update with Git.

License

This project is released under the Apache License 2.0. See LICENSE.

Contact

Maintained by Zijiang Song. Contact: jacksung1995@gmail.com.