Parse description string to populate NULL fields - parsing

Question

Informix 12.10
tblItems
(
Type SMALLINT, {Precious Metal = 1, Other = 2}
Description VARCHAR,
Quantity SMALLINT,
Name VARCHAR,
Weight DECIMAL(5,1),
Purity SMALLINT,
Brand VARCHAR,
Model VARCHAR,
SerialNum VARCHAR
);
EDIT UPDATE: Sample data below is stored in tblItems.Type and tblItems.Description. Please note that the contents in Description column are all uppercase characters and may also include punctuation character.
2|1LAPTOP APPLE 15.5" MODEL MACKBOOK PRO,S/N W80461WCAGX, WITH CHARGER||||||||
1|1RING 2.3PW 14K||||||||
2|DRILL RIOBY, MODEL D5521 S/N77720||||||||
2|TRIMMER TORO, MODEL 0242 S/N 66759||||||||
2|CELL SAMSUNG NOTE3, MODEL SM-N900T S/N RV8F90YLZ9W||||||||
I need to parse the sample item descriptions into the columns below, using the rules mentioned in the comments :
Quantity, {if description string does not start with a number, then Quantity = 1}
Name, {Always the first element if description has no quantity, second element if quantity present]
Weight, {Always before "PW" if Type = 1, Default to zero if Type = 2}
Purity, {Always before "K" if Type = 1, Default to NULL if Type = 2}
Brand, {Always the second element in description, if present}
Model, {Always after "MODEL", with or without a space}
Serial Number {Always after "S/N", with or without a space}
I would like to do this with an UPDATE statement, but if Informix has an import utility tool like SQL-Server's SSIS, then that could be a better option.
UPDATE, Expected Results:
Quantity 1 1 1 1 1
Name LAPTOP RING DRILL TRIMMER CELL
Weight 0.0 2.3 0.0 0.0 0.0
Purity 14
Brand APPLE RIOBY TORO SAMSUNG
Model MACKBOOK PRO D5521 0242 SM-N900T
SerialNum W8046WCAGX 77720 66759 RV8F90YLZ9W

Answer 1

Assuming you are using Informix 12.10.XC8 or above, you can try using regular expressions to parse the description string (see the online documentation here).
For the serial number, for example, you can do:
UPDATE tblitems
SET
serialnum =
DECODE
(
regex_match(description, '(.*)(S\/N)(.*)', 3)
, 't'::BOOLEAN, regex_replace(description, '(.*)(S\/N)([[:blank:]]?)([[:alnum:]]*)(.*)', '\4', 0, 3)
, 'f'::BOOLEAN, ''
)
So in the previous example I am testing if the description contains the S/N string and if that is true I use regex_replace to return the value after it, in this case the 4th matching group in the regular expression (I am not using regex_extract to get the value because it seems to return multiple values and I get error -686).
You can extend this approach to the rest of the columns and see if regular expressions are enough to parse the description column.

Answer 2

If you're looking for a SQL Server option and open to a Split/Parse function which maintains the sequence
Example
Select A.Type
,A.Description
,C.*
From YourTable A
Cross Apply (values ( replace(
replace(
replace(
replace(A.Description,',',' ')
,' ',' ')
,'Model ','Model')
,'S/N ','S/N')
)
)B(CleanString)
Cross Apply (
Select Quantity = IsNull(left(max(case when RetSeq=1 then RetVal end),NullIf(patindex('%[^0-9]%',max(case when RetSeq=1 then RetVal end)) -1,0)),1)
,Name = substring(max(case when RetSeq=1 then RetVal end),patindex('%[^0-9]%',max(case when RetSeq=1 then RetVal end)),charindex(' ',max(case when RetSeq=1 then RetVal end)+' ')-1)
,Weight = IIF(A.Type=2,null,try_convert(decimal(5,1),replace(max(case when RetVal like '%PW' then RetVal end),'PW','')))
,Purity = try_convert(smallint ,replace(max(case when RetVal like '%K' then RetVal end),'K',''))
,Brand = IIF(A.Type=1,null,max(case when RetSeq=2 then RetVal end))
,Model = replace(max(case when RetVal Like 'Model[0-9,A-Z]%' then RetVal end),'Model','')
,SerialNum = replace(max(case when RetVal Like 'S/N[0-9,A-Z]%' then RetVal end),'S/N','')
From [dbo].[tvf-Str-Parse](CleanString,' ') B1
) C
Returns
The TVF if Interested
CREATE FUNCTION [dbo].[tvf-Str-Parse] (#String varchar(max),#Delimiter varchar(10))
Returns Table
As
Return (
Select RetSeq = Row_Number() over (Order By (Select null))
,RetVal = LTrim(RTrim(B.i.value('(./text())[1]', 'varchar(max)')))
From (Select x = Cast('<x>' + replace((Select replace(#String,#Delimiter,'§§Split§§') as [*] For XML Path('')),'§§Split§§','</x><x>')+'</x>' as xml).query('.')) as A
Cross Apply x.nodes('x') AS B(i)
);
EDIT - If you don't want or can't use a TVF
dbFiddle
Select A.Type
,A.Description
,C.*
From YourTable A
Cross Apply (values ( replace(
replace(
replace(
replace(A.Description,',',' ')
,' ',' ')
,'Model ','Model')
,'S/N ','S/N')
)
)B(CleanString)
Cross Apply (
Select Quantity = IsNull(left(max(case when RetSeq=1 then RetVal end),NullIf(patindex('%[^0-9]%',max(case when RetSeq=1 then RetVal end)) -1,0)),1)
,Name = substring(max(case when RetSeq=1 then RetVal end),patindex('%[^0-9]%',max(case when RetSeq=1 then RetVal end)),charindex(' ',max(case when RetSeq=1 then RetVal end)+' ')-1)
,Weight = IIF(A.Type=2,null,try_convert(decimal(5,1),replace(max(case when RetVal like '%PW' then RetVal end),'PW','')))
,Purity = try_convert(smallint ,replace(max(case when RetVal like '%K' then RetVal end),'K',''))
,Brand = IIF(A.Type=1,null,max(case when RetSeq=2 then RetVal end))
,Model = replace(max(case when RetVal Like 'Model[0-9,A-Z]%' then RetVal end),'Model','')
,SerialNum = replace(max(case when RetVal Like 'S/N[0-9,A-Z]%' then RetVal end),'S/N','')
From (
Select RetSeq = row_number() over (Order By (Select null))
,RetVal = ltrim(rtrim(B.i.value('(./text())[1]', 'varchar(max)')))
From (Select x = Cast('<x>' + replace((Select replace(CleanString,' ','§§Split§§') as [*] For XML Path('')),'§§Split§§','</x><x>')+'</x>' as xml).query('.')) as A
Cross Apply x.nodes('x') AS B(i)
) B1
) C