ryayoung/scopespace

A simple Python context manager for creating scoped namespaces. Great for organizing code and versioning dataframes in notebooks.

Scopespace

A new design pattern for working with data in a notebook environment.

pip install scopespace

Quickstart: Learn by example

ScopeSpace is a context manager whose with block has its own local scope.

x = 5
with ScopeSpace() as foo:
    x = x + 1
    
print(x)  # 5
print(foo.x) # 6

Notice we did not overwrite x globally. Instead, we have foo, a namespace with it's own version of x:

Anything we declare within a ScopeSpace is isolated within our chosen namespace.

with ScopeSpace() as bar:
    stuff = 10
    
print(stuff)  # NameError: name 'stuff' is not defined
print(bar.stuff)  # 10

What's the point?

A common challenge for data people

In the world of dataframes and notebook environments, naming things can be tough. You're often juggling multiple references to the same underlying data. Describing one of those with a variable name is difficult. And you're often forced to version your data by changing its name across cells, to make each cell idempotent. This confuses the reader.

There's a conflict of interest here:

• While the following may be true ...
  1. You want to use descriptive names.
  2. You want to incorporate versioning/renaming so cells can be run repeatedly.
• There are a few big problems with this:
  1. Renaming your data 10 times can make your code seem far more complex than it really is.
  2. In some libraries, like Pandas, you're strongly incentivized, naturally, to use short names, since names are often used repeatedly in the same command, to access columns.

A Typical Example

You've just created a new notebook, and have quickly jotted down some code to perform a few manipulations on a Pandas dataframe, displaying the data at each stage.

2 problems arise from this code:

1. The second and third cells will error if we try to run either of them twice in a row.
2. The name, df is not very descriptive.

So we switch to more descriptive names, and version them across cells.

Now we've created a few more problems:

1. Our code is verbose and redundant. The second line in the third cell now takes 92 characters (the name of the dataframe occupies 57 of them), just to express c = (c + b).astype(str).
2. It's not clear to the reader that we are only working with one dataset.
3. We've created room for mistakes.

A common solution is to use functions with descriptive names and a df parameter. This may present its own issues though:

1. Introduces several unnecessary steps: 1. Declare function signature. 2. Return something. 3. Call it, and pass it arguments.
2. We never intended to use it anywhere except immediately after declaration. Thus, its purpose is unclear to the reader.

ScopeSpace: An alternative

"Namespaces are one honking great idea -- let's do more of those!" - Tim Peters, 'The Zen of Python'

Here, we took the idea of a notebook cell and gave it a logical structure - an improvement in both form, and function.

By isolating logical tasks to their own scopes:

1. Each cell is expressive of its own purpose; the namespace labels are explicitly declared first
2. We made our naming style consistent, with an obvious, repeatable pattern
3. The code is more concise, but without sacrificing details
4. We eliminated redundancy: Descriptive names (the namespace labels) no longer clutter our logic when working with the data.

By grouping similar tasks together under a common namespace:

1. We get the logical grouping benefit that functions provide, but without needing to define or call one.
2. We get the attribute organization benefit that class instances provide, but without having to define our own class.