Complicated Comments

A "Question of the Day" I submitted to SQLServerCentral.com was published yesterday, and it seemed to generate a lot of interest, due to the rather surprising answer.

The question was, "Given the code below, which numbers will be printed?"

PRINT '1' -- /* ;PRINT '2' */ ;PRINT '3' /*
PRINT '4' --*/
--/*
PRINT '5'
--*/
/*
PRINT '6'
--/*
*/
PRINT '7'
--*/
PRINT '8'

Possible answers:
A. 1, 8
B. 1, 4, 8
C. 1, 4, 5, 8
D. 1, 4, 5, 7, 8

Correct answer: C

Fewer than half of the respondents answered correctly (I certainly would have gotten the question wrong myself - this question derived from research that completely surprised me). It's not exactly a trick question, but the answer centers around a little-known feature of nested comments. It's fairly obvious that the numbers 1, 4, 5, and 8 will print. The tricky part is near the bottom of the code. Most people look at the block comment starting above the line "PRINT '6'" and think that it ends above the line "PRINT '7'", following the reasoning that the line below "PRINT '6'", "--/*" is ignored by the SQL parser. But because the inline comment of that line is already inside a block comment, the inline comment is ignored! Which means that the line "--/*" actually begins a new, nested block comment that closes on the next line, but causes the line "PRINT '7'" to remain commented-out. In other words, the block comment containing "PRINT '6'" also contains "PRINT '7'", and does not close until the line above "PRINT '8'". This confusion explains why, as of right now, 40% of the respondents chose answer "D".

When I submitted the question for publication, I wondered that many people would find it too abstruse and esoteric to be of value. A number of participants in the discussion of the question pointed out that if they saw comments similar to this in a code review, they would send the developer back to his or her desk to clean up. And point taken. Nesting comments in this fashion has very little, if any, upside and almost 100% downside in that it can lead to largely unexpected results. A couple of articles on the site help to establish best practices for commenting: "The case against using single-line comments", by Cade Bryant, and "Worst Practice - Bad Comments", by Andy Warren.

This problem originally came to mind when I started work on a tool for comparing the text from stored procedures. I wanted to allow the option of omitting all comments from the comparison, which of course would require me to identify and parse out comments from the text. In my research, I came across the question of how to treat nested comments - the "order of operation" of inline versus block comments, nested blocks, etc., and that's when I came across the scenario that formed this question of the day. I have to say that I was surprised, almost to the point of shock, that commenting functioned this way.