The Till Export

A chain of market stalls sends its sales to head office as raw CSV text — one long string, straight out of each stall's till. Before anyone can add anything up, that text has to be turned back into rows.

Task: write totals_by(text, key, value). It takes the CSV text, the name of a column to group by (key) and the name of a column of numbers to add up (value). It returns a dictionary mapping each key value to the total of its value column. The order of the keys doesn't matter.

How the text is laid out

• Lines are separated by "\n", and the first line is always the header row of column names.
• Every till writes its columns in its own order, so a column has to be found by its name, never by its position.
• No value contains a comma. Values and column names can have stray spaces around them: " north " means north.

Tills are messy. Skip a line — quietly, without crashing — when:

• it is blank (the text usually ends with "\n", which leaves an empty last line),
• it has a different number of values from the header,
• its key is empty, or its value is empty or the text NA.

Every other value is a valid number, possibly with a decimal point. Keys stay as text, even when they look like numbers.

Example

The NA line is skipped, and so is the empty line after the final "\n".