babypandas

Very quick hack, making a pure python implementation of Pandas' DataFrame/Series (very incomplete and hacky!!)

Note currently doesn't have index support. Just the very, very, very basics.

Hi, I wrote this as I need to rewrite some code to use pure python, instead of numpy/Pandas for performance reasons. (working with lots of tiny datasets, not a few medium/big ones). And I really missed the convenience of the DataFrame/Series datastructures.

Working with a list of dictionaries just isn't the same.

Found having to call dict constructor in comprehensions to do what I wanted.

e.g.

    data = [
        {'a': 1, 'b': 'banana', 'c': 33},
        {'a': 3, 'b': 'cat', 'c': 3},
    ]
    df = DataFrame(data)

    # add a column (scalar)
    data =[dict(row, d=3) for row in data]
    # vs
    df['d'] = 3

    # add a column (list-like)
    vals = [1, 2]
    data = [dict(row, d=v) for row, v in zip(data, vals)]
    # vs
    df['d'] = vals

    # addition of columns
    data = [dict(row, sum=(row['a'] + row['c'])) for row in data]
    # vs
    df['sum'] = df['a'] + df['c']
    
    # map
    data = [dict(row, small=str.lower(row['b])) for row in data]
    # vs
    df['small'] = df['b'].map(str.lower)
    # or
    df['small'] = df['b'].map(lambda x: x.lower())

    # filter
    data = [row for row in data if row['a'] > 1]
    # vs
    df = df[df['a'] > 1]

    # percentage
    tot = sum(row['a'] for row in data)
    data = [dict(row, pct=(100.0 * row[a'] / tot) for row in data]
    # vs
    df['pct'] = 100.0 * df['a'] * df['a'].sum()