Step 3: handle bad input

Step 2 treats anything that isn't the right number as a wrong answer, including a typo. This step asks again instead, and lets an empty answer skip the question. The question's body becomes a loop that ends when the answer has been dealt with:

while (true)
{
    Out.Write($"Q{i}: {a} + {b} = ");
    int length = In.ReadLine(buffer);
    if (length < 0)
    {
        Out.Line();
        Out.Line("(end of input)");
        asked--;
        goto Done;
    }
    // A line longer than the buffer returns its full length; the
    // buffer holds only its start, so ask again.
    if (length > buffer.Length)
    {
        Out.Write($"that's {length} characters; ");
        Out.Line($"answers are at most {buffer.Length}");
        continue;
    }

    ReadOnlySpan<char> answer = buffer.Slice(0, length).Trim();
    if (answer.Length == 0)
    {
        Out.Line("skipped");
        break;
    }
    if (!Str.TryParse(answer, out long value))
    {
        Out.Line($"\"{answer}\" isn't a number");
        continue;
    }

    if (value == a + b)
    {
        right++;
        Out.Line("right");
    }
    else
    {
        Out.Line($"no, it's {a + b}");
    }
    break;
}

Three cases are new:

  • Too long. In.ReadLine returns the line's full length even when the buffer can't hold it, and skips the rest of the line, so the next read starts on the next line. length > buffer.Length is how the program knows. Returning the length instead of failing lets the caller decide: here, ask again.
  • Empty. After Trim(), an empty span means the answer was blank: skip the question.
  • Not a number. Str.TryParse returns false. The message quotes the answer with {answer}: a span can be a hole in $"...", so the text goes from the input buffer to the output buffer without a string in between.

goto Done leaves both loops at the end of input. C#'s goto is as unfashionable as ever, but it is the plainest way out of two loops.

Run it

The test suite feeds this step a file with every kind of bad input: letters, characters outside ASCII, bytes that aren't valid UTF-8, a 40-character line, a Windows line ending, an empty line, and a last line without a line ending.

./build.sh QuizStep3
./run.sh QuizStep3 < tests/input/QuizStep3.txt
6 questions. An empty answer skips one.
Q1: 38 + 8 = "abc" isn't a number
Q1: 38 + 8 = right
Q2: 5 + 0 = "é😀" isn't a number
Q2: 5 + 0 = "��(" isn't a number
Q2: 5 + 0 = no, it's 5
Q3: 12 + 5 = that's 40 characters; answers are at most 32
Q3: 12 + 5 = right
Q4: 2 + 43 = skipped
Q5: 27 + 8 = right
Q6: 38 + 36 = 
(end of input)
3 of 5 right

The invalid bytes came out as �, the Unicode replacement character: In decodes UTF-8 and replaces what it can't decode, so a program never sees a broken character. "é😀" went in as UTF-8, became UTF-16 chars in the span, and came back out as UTF-8 through Out.

Next: keep the mistakes.