Question

I encountered a bit of code within the book I’ve been reading that has me questioning the SUBSTRING() function’s behavior. The code is supposed to search a NYSIIS Replacement table (phonetic encoding example) and replace the middle ‘N-gram’ of an input string based on the location ‘End’ ‘Mid’ or ‘Start’ in the table. an excerpt is provided below:

NYSIIS Replacement Table:

Location NGram Replacement

Mid      A     A
Mid      AW    AA
Mid      E     A
Mid      EV    AF
Mid      EW    AA
Mid      I     A

USE [AdventureWorks]

DECLARE @Result NVARCHAR(100) = N'NEVADA';

DECLARE @Replacement NVARCHAR(10);

DECLARE @i INT;

SET @i = 1;

WHILE @i <= LEN (@Result)

BEGIN

    SET @Replacement = NULL;

    -- Grab the middle-of-name replacement n-gram

    SELECT TOP(1)  @Replacement = Replacement                   
    FROM dbo.NYSIIS_Replacements                         
    WHERE Location = N'Mid'
        AND SUBSTRING(@Result, @i, LEN(NGram)) = NGram
    ORDER BY LEN(NGram) DESC;


    SET @Replacement = COALESCE(@Replacement, SUBSTRING(@Result, @i, 1));   


    -- If we found a replacement, apply it

    SET @Result = STUFF(@Result, @i, LEN(@Replacement), @Replacement) 

    -- Move on to the next n-gram

    SET @i = @i + COALESCE(LEN(@Replacement), 1);


END;

SELECT @Result;

When the SUBSTRING() function encounters 2 possible matches using ‘NEVADA’ as an example (‘E’ and ‘EV’ in the table) how does it ‘know’ to use the 2 letter string as opposed to the one? Is this the expected behavior for SUBSTRING()?

I would assume the @Replacement variable would contain both ‘A’ and ‘AF’ but when debugging it only appears to contain ‘N’ in the first iteration and ‘AF’ in the second.

Also I could not understand why TOP and ORDER BY were included in this example. Commenting them out produces the same results.

Answer 1

The ORDER BY clause uses the length of the pattern and sorts in descending order, hence the longest match will occur first. The TOP clause limits the results to the first row. Removing the ORDER BY clause makes the result unpredictable.

COALESCE is used to set @Replacement to either the replacement pattern, or the character at position @i in the @Result string if no pattern match was found.