Data science: Profile every column

A profile is a one-line summary of every column, the first output of almost every analysis. Decide each column's kind and summarise it.

What to build

A program that profiles each column of a CSV table.

Requirements

  • Empty cells are missing and are ignored in the summaries. If every cell of a column is empty, print NAME: empty.
  • If every non-empty cell is a number, print NAME: numeric, missing M, min A, max B, mean C, with A and B in :g format and C to two decimals.
  • Otherwise print NAME: text, missing M, U unique, top T (K): the most common value and its count, taking the alphabetically first value on a tie.

Input

Line 1: n. Line 2: the header. Then n rows, cells separated by commas.

Output

One line per column, in order.

Constraints

1 ≤ n ≤ 1000; no cell contains a comma; numbers are between −100000 and 100000.

Example

Input:

5
city,temp,sky
Oslo,12.6,cloudy
Rome,24,sunny
Oslo,,rain
Paris,18,cloudy
Rome,21,sunny

Output:

city: text, missing 0, 3 unique, top Oslo (2)
temp: numeric, missing 1, min 12.6, max 24, mean 18.90
sky: text, missing 0, 3 unique, top cloudy (2)

Why it works

Trying float() on every value is a reliable test of a numeric column, and the summaries that fit a column depend on its kind: range and mean for numbers, distinct values and the most common one for text.

Examples

Example 1 · Input
5
city,temp,sky
Oslo,12.6,cloudy
Rome,24,sunny
Oslo,,rain
Paris,18,cloudy
Rome,21,sunny
Expected output
city: text, missing 0, 3 unique, top Oslo (2)
temp: numeric, missing 1, min 12.6, max 24, mean 18.90
sky: text, missing 0, 3 unique, top cloudy (2)
Example 2 · Input
3
id,score
1,90
2,85
3,95
Expected output
id: numeric, missing 0, min 1, max 3, mean 2.00
score: numeric, missing 0, min 85, max 95, mean 90.00

Prompts inside input() are ignored when checking, so input("n: ") is fine.

Hints

Hint 1

A column is numeric when every non-empty cell can be converted with float(); try it inside try/except ValueError.

Hint 2

For text columns, count values with a dictionary and pick the top one with min(counts, key=lambda v: (-counts[v], v)).

Hint 3

f"{x:g}" prints 24.0 as 24 and 12.6 as 12.6.

Solution

Show one correct solution (try it yourself first)
n = int(input())
header = input().split(",")
rows = [input().split(",") for _ in range(n)]
for i, name in enumerate(header):
    values = [r[i] for r in rows if r[i] != ""]
    missing = n - len(values)
    if not values:
        print(f"{name}: empty")
        continue
    try:
        nums = [float(v) for v in values]
    except ValueError:
        nums = None
    if nums is not None:
        mean = sum(nums) / len(nums)
        print(f"{name}: numeric, missing {missing}, min {min(nums):g}, max {max(nums):g}, mean {mean:.2f}")
    else:
        counts = {}
        for v in values:
            counts[v] = counts.get(v, 0) + 1
        top = min(counts, key=lambda v: (-counts[v], v))
        print(f"{name}: text, missing {missing}, {len(counts)} unique, top {top} ({counts[top]})")

Lesson: Exploratory data analysis →