Converting a table with a key and comment field into a key and row for every word in the column field - tsql

Question

I have a table with unstructured data I am trying to analyze to try to build a relational lookup. I do not have use of word cloud software.
I really have no idea how to solve this problem. Searching for solutions has lead me to tools that might do this for me that cost money, not coded solutions.
Basically my data looks like this:
CK1 CK2 Comment
--------------------------------------------------------------
1 A This is a comment.
2 A Another comment here.
And this is what I need to create:
CK1 CK2 Words
--------------------------------------------------------------
1 A This
1 A is
1 A a
1 A comment.
2 A Another
2 A comment
2 A here.

Answer 1

What you are trying to do is tokenize a string using a space as a Delimiter. In the SQL world people often refer to functions that do this as a "Splitter". The potential pitfall of using a splitter for this type of thing is how words can be separated by multiple spaces, tabs, CHAR(10)'s, CHAR(13)'s, CHAR()'s, etc. Poor grammar, such as not adding a space after a period results in this:
" End of sentence.Next sentence"
sentence.Next is returned as a word.
The way I like to tokenize human text is to:
Replace any text that isn't a character with a space
Replace duplicate spaces
Trim the string
Split the newly transformed string using a space as the delimiter.
Below is my solution followed by the DDL to create the functions used.
-- Sample Data
DECLARE #yourtable TABLE (CK1 INT, CK2 CHAR(1), Comment VARCHAR(8000));
INSERT #yourtable (CK1, CK2, Comment)
VALUES
(1,'A','This is a typical comment...Follewed by another...'),
(2,'A','This comment has double spaces and tabs and even carriage
returns!');
-- Solution
SELECT t.CK1, t.CK2, split.itemNumber, split.itemIndex, split.itemLength, split.item
FROM #yourtable AS t
CROSS APPLY samd.patReplace(t.Comment,'[^a-zA-Z ]',' ') AS c1
CROSS APPLY samd.removeDupChar8K(c1.newString,' ') AS c2
CROSS APPLY samd.delimitedSplitAB8K(LTRIM(RTRIM(c2.NewString)),' ') AS split;
Results (truncated for brevity):
CK1 CK2 itemNumber itemIndex itemLength item
----------- ---- -------------------- ----------- ----------- --------------
1 A 1 1 4 This
1 A 2 6 2 is
1 A 3 9 1 a
1 A 4 11 7 typical
1 A 5 19 7 comment
...
2 A 1 1 4 This
2 A 2 6 7 comment
2 A 3 14 3 has
2 A 4 18 6 double
...
Note that the splitter I'm using is based of Jeff Moden's Delimited Split8K with a couple tweeks.
Functions used:
CREATE FUNCTION dbo.rangeAB
(
#low bigint,
#high bigint,
#gap bigint,
#row1 bit
)
RETURNS TABLE WITH SCHEMABINDING AS RETURN
WITH L1(N) AS
(
SELECT 1
FROM (VALUES
(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),
(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),
(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),
(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),(0),
(0),(0)) T(N) -- 90 values
),
L2(N) AS (SELECT 1 FROM L1 a CROSS JOIN L1 b CROSS JOIN L1 c),
iTally AS (SELECT rn = ROW_NUMBER() OVER (ORDER BY (SELECT 1)) FROM L2 a CROSS JOIN L2 b)
SELECT r.RN, r.OP, r.N1, r.N2
FROM
(
SELECT
RN = 0,
OP = (#high-#low)/#gap,
N1 = #low,
N2 = #gap+#low
WHERE #row1 = 0
UNION ALL -- COALESCE required in the TOP statement below for error handling purposes
SELECT TOP (ABS((COALESCE(#high,0)-COALESCE(#low,0))/COALESCE(#gap,0)+COALESCE(#row1,1)))
RN = i.rn,
OP = (#high-#low)/#gap+(2*#row1)-i.rn,
N1 = (i.rn-#row1)*#gap+#low,
N2 = (i.rn-(#row1-1))*#gap+#low
FROM iTally AS i
ORDER BY i.rn
) AS r
WHERE #high&#low&#gap&#row1 IS NOT NULL AND #high >= #low AND #gap > 0;
GO
CREATE FUNCTION samd.NGrams8k
(
#string VARCHAR(8000), -- Input string
#N INT -- requested token size
)
RETURNS TABLE WITH SCHEMABINDING AS RETURN
SELECT
position = r.RN,
token = SUBSTRING(#string, CHECKSUM(r.RN), #N)
FROM dbo.rangeAB(1, LEN(#string)+1-#N,1,1) AS r
WHERE #N > 0 AND #N <= LEN(#string);
GO
CREATE FUNCTION samd.patReplace8K
(
#string VARCHAR(8000),
#pattern VARCHAR(50),
#replace VARCHAR(20)
)
RETURNS TABLE WITH SCHEMABINDING AS RETURN
SELECT newString =
(
SELECT CASE WHEN #string = CAST('' AS VARCHAR(8000)) THEN CAST('' AS VARCHAR(8000))
WHEN #pattern+#replace+#string IS NOT NULL THEN
CASE WHEN PATINDEX(#pattern,token COLLATE Latin1_General_BIN)=0
THEN ng.token ELSE #replace END END
FROM samd.NGrams8K(#string, 1) AS ng
ORDER BY ng.position
FOR XML PATH(''),TYPE
).value('text()[1]', 'VARCHAR(8000)');
GO
CREATE FUNCTION samd.delimitedSplitAB8K
(
#string VARCHAR(8000), -- input string
#delimiter CHAR(1) -- delimiter
)
RETURNS TABLE WITH SCHEMABINDING AS RETURN
SELECT
itemNumber = ROW_NUMBER() OVER (ORDER BY d.p),
itemIndex = CHECKSUM(ISNULL(NULLIF(d.p+1, 0),1)),
itemLength = CHECKSUM(item.ln),
item = SUBSTRING(#string, d.p+1, item.ln)
FROM (VALUES (DATALENGTH(#string))) AS l(s) -- length of the string
CROSS APPLY
(
SELECT 0 UNION ALL -- for handling leading delimiters
SELECT ng.position
FROM samd.NGrams8K(#string, 1) AS ng
WHERE token = #delimiter
) AS d(p) -- delimiter.position
CROSS APPLY (VALUES( --LEAD(d.p, 1, l.s+l.d) OVER (ORDER BY d.p) - (d.p+l.d)
ISNULL(NULLIF(CHARINDEX(#delimiter,#string,d.p+1),0)-(d.p+1), l.s-d.p))) AS item(ln);
GO
CREATE FUNCTION dbo.RemoveDupChar8K(#string varchar(8000), #char char(1))
RETURNS TABLE WITH SCHEMABINDING AS RETURN
SELECT NewString =
replace(replace(replace(replace(replace(replace(replace(
#string COLLATE LATIN1_GENERAL_BIN,
replicate(#char,33), #char), --33
replicate(#char,17), #char), --17
replicate(#char,9 ), #char), -- 9
replicate(#char,5 ), #char), -- 5
replicate(#char,3 ), #char), -- 3
replicate(#char,2 ), #char), -- 2
replicate(#char,2 ), #char); -- 2
GO

Answer 2

1) If we are using SQL Server 2016 and above then we should probably
use the built-in function STRING_SPLIT
-- SQL 2016and above
DECLARE #txt NVARCHAR(100) = N'This is a comment.'
select [value] from STRING_SPLIT(#txt, ' ')
2) Only if 1 does not fit, then if the number of separation (the space in our case) is less then 3 which fit your sample data, then we should probably use PARSENAME
-- BEFORE SQL 2016 if we have less than 4 parts
DECLARE #txt NVARCHAR(100) = N'This is a comment.'
DECLARE #Temp NVARCHAR(200) = REPLACE (#txt,'.','#')
SELECT t FROM (VALUES(1),(2),(3),(4))T1(n)
CROSS APPLY (SELECT REPLACE(PARSENAME(REPLACE(#Temp,' ','.'),T1.n), '#','.'))T2(t)
3) Only if the 1 and 2 does not fit, then we should use SQLCLR function
http://dataeducation.com/sqlclr-string-splitting-part-2-even-faster-even-more-scalable/
4) Only if we cannot use 1,2 and we cannot use SQLCLR (which implies a real problematic administration and has nothing with security since you can have all the SQLCLR function in a read-only database for the use of all users, as I explain in my lectures), then you can use T-SQL and create UDF.
https://sqlperformance.com/2012/07/t-sql-queries/split-strings