# 8A inspection checkpoint. Run before completing 08_pandas.py.
from io import StringIO
import pandas as pd
text = "topic,minutes\n arrays ,15\ntables,not yet\ntables,0\n charts ,-4\n,20\n"
raw = pd.read_csv(StringIO(text), dtype="string", keep_default_na=False)
# First inspect shape, head, dtypes, empty strings and isna counts.
# Record THREE observations and the action each suggests before changing raw.
observations = ["TODO: observation and action", "TODO", "TODO"]
# Predict: how many numeric conversions fail? How many negative values?
# How many empty normalized topics? Could those failure sets overlap?
# Then make a working copy, convert/normalize and compare predictions.
# Finally state the accepted labels, not only the number of accepted rows.
