I keep running across comments and answers to questions that imply that the mapping cylinder for “nice” functions is a CW complex. Why is this necessarily so?
Consider any function $ f \colon X \to Y $. Using Hatcher’s (p2) standard definition, the mapping cylinder will not be a CW complex unless Y, the range of f, is a CW complex.
If f is a nice function, its image may be nice, but the rest of Y can be almost anything.
What am I missing?