forgebox 0.3.14

Data science with fancy loops

ForgeBox

Data science comprehensive toolbox 🛠⚔️📦

Installation

Easy simple installation in 1 shot

pip install forgebox

If not specified, you need anaconda3 for most of the tools, python shold be at least >=3.6

Features 🚀 Briefing

This is a tool box with comprehensive utilies, to put it simply, I just hope most of my frequetyly used DIY tools in in place and can be easily installed and imported

Lazy, fast imports 🤯

The following command will import many frequent tools for data science, like pd for pandas, np for numpy, os, json, PIL.Image for image processing

from frogebox.imports import *

No more🚫 following typings

import pandas as pd
import numpy as np
import os
import json
...

Free style mapping

Works on every value of a complicated dictionary structure (eg. list in dict in list in dict, etc,. 😳)

from forgebox.freemap import FreeMap

# flatten decides if we want to flatten the strucuture
freemap_tool = FreeMap(
    <function/callable applying to every value>,
    <function/callable that filters every value>,
    flatten=True
)

data2 = freemap_tool(data1)

Interactive Widgets

Interactive widgets work with in jupyter notebooks

Search box 🔎 for dataframe

This will create an interactive text input box to search through the pandas dataframe, within the columns you set.

if manual is set to False, the search will respond to each of your key press, it's fast but will suffer terrible user experience if the dataframe is huge in size.

from forgebox.widgets import search_box

search_box(data_df, columns=["col1","col2"], manual=False)

paginate

You can browse through a pandas dataframe like fliping pages 📄.

from forgebox.widgets import paginate

paginate(your_dataframe, page_len=10)

Single button callback

a fully functional page with a single button, this single button is bonded to a function

This is as much code as you need, to build a fully functional interactive page shows sql table from jupyter, that you can:✅ choose which table to visit✅ choose how many lines you want to show, (with a slider)✅ configure the where condition with a text box on front end

tablename_list = ["pubmed", "patient", "users", "drugs"]

from forgebox.html import DOM
def show_sql_table(sql_input:str) -> str:
    with engine.connect() as conn:
        df=pd.read_sql(sql_input, con=conn)
    # display the table as html
    DOM(df.to_html(),"div")()

@SingleButton(callback=show_sql_table)
def abc(
    limit:{"typing":int, "default":10, "min":5, "max":20},
    where_condition:{"typing":str, "default": "where 1=1", },
    table:{"typing":list, "options":tablename_list}
):
    return f"sql > SELECT * FROM {table} {where_condition} LIMIT {limit}"

Dataset Layering 🍰

Instead of creating a bunch of dataset just treat another dataset as extra layer of getitem function

@layering(SomeStringDataset, "tensorDataset")
def tokenizing(x):
    return tokenizer(x, return_tensors='pt')['input_ids'][0]

@layering(tokenizing, "guessNextDataset")
def guess_next(x):
    return x[:-1], x[1:]

some_string_data = SomeStringDataset()

tokenized_data = some_string_data.next_layer()
print(tokenized_data[3])

guess_pair = tokenized_data.next_layer()
print(guess_pair[3])

And you can test these layer functions one by one