topic Table transformations in QlikView

Table transformations

Thu, 11 Sep 2014 13:28:12 GMT

Hi I have table with one column with structure as is here:

I’d like to apply some transformation, but I don’t know which exactly.

Info between two signs ‘[group]’ is one entry, with 3 columns: gid, user, role (there are exceptions where only gid is available). User and role might have several entries.

My aim is to have at least this table:

gid	user	role
g1	u1,u2,u3	r1,r2,r3
g2
g3	u1	r1

Please help me with some hints.

Re: Table transformations

Thu, 11 Sep 2014 13:42:33 GMT

Maybe subfield()?

subfield(s, 'delimiter' [ , index ] )

In its three-parameter version, this script function returns a given substring from a larger string s with delimiter 'delimiter'. index is an optional integer denoting which of the substrings should be returned. If index is omitted when subfield is used in a field expression in a load statement, the subfield function will cause the load statement to automatically generate one full record of input data for each substring that can be found in s.

In its two-parameter version, the subfield function generates one record for each substring that can be taken from a larger string s with the delimiter 'delimiter'. If several subfield functions are used in the same load statement, the Cartesian product of all combinations will be generated.

Examples:

(For three parameters)

subfield(S, ';' ,2) returns 'cde' if S is 'abc;cde;efg'

subfield(S, ';' ,1) returns NULL if S is an empty string

subfield(S, ';' ,1) returns an empty string if S is ';'

Re: Table transformations

rubenmarin — Thu, 11 Sep 2014 13:57:41 GMT

Hi Dusan,

To create a table from that only column you can use something like:

LOAD mid(@1, 1, Index(@1, '=')-1) as field,

mid(@1, Index(@1, '=')+1) as value,

If(mid(@1, 1, Index(@1, '=')-1)='guid', NumSum(1, Peek('Cont')), Peek('Cont')) as Cont

Resident temp where @1<>'[group]';

Re: Table transformations

marcus_sommer — Thu, 11 Sep 2014 15:00:35 GMT

There is a similar data-structure after the export from windows event-logs to txt-files per "wevtutil". This could be structured per mod(rowno()) - here an example:

... some loops for all logs

... some prepare and cleaning

PrepareData:

Load

EventNr, mod(rowno(), 12) as Mod, /*LineString,*/ LogSource,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 11, 'RawDataResult')) as LogName,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 10, 'RawDataResult')) as Source,

if(mod(rowno(), 12) = 0, date(num(mid(Peek('Value', rowno() - 1 - 9, 'RawDataResult'), 1, 10)) +

num(mid(Peek('Value', rowno() - 1 - 9, 'RawDataResult'), 12, 8)), 'DD.MM.YYYY hh:mm:ss')) as Date,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 8, 'RawDataResult')) as EventID,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 7, 'RawDataResult')) as Task,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 6, 'RawDataResult')) as Level,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 5, 'RawDataResult')) as Opcode,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 4, 'RawDataResult')) as Keyword,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 3, 'RawDataResult')) as User,

if(mod(rowno(), 12) = 0, Peek('Value', rowno() - 1 - 2, 'RawDataResult')) as UserName,

if(mod(rowno(), 12) = 0, peek('Value', rowno() - 1 - 1, 'RawDataResult')) as Computer,

if(mod(rowno(), 12) = 0, Value) as Description

Resident RawDataResult;

SelectDataSystem:

Noconcatenate Load * Resident PrepareData Where Mod = 0;

... some cluster ...

I think you could adapt this approach.

- Marcus

Re: Table transformations

Thu, 11 Sep 2014 16:34:13 GMT

Thank you for ideas.

It's maybe not clear from the screenshot, but the text consists of several rows, so the first row is [group], 2nd row 'gid=...', 3rd row is 'user=...' and so on. So subfield() and mid() are not suitable so far.

@Marcus I'll need some time to modify my data structure on your script.

Re: Table transformations

maxgro — Thu, 11 Sep 2014 20:47:20 GMT

see attachment

Re: Table transformations

MarcoWedel — Thu, 11 Sep 2014 22:45:43 GMT

Hi,

one possible solution:

Input:

Output:

used script:

tabInput:
LOAD NumSum(Peek(ID),-(Text='[group]')) as ID,
    SubField(Text,'=',1) as ColNam,
    SubField(Text,'=',2) as ColVal;
LOAD [@1:n] as Text
FROM [http://community.qlik.com/servlet/JiveServlet/download/607776-124887/QlikCommunity_Thread_132964.txt] (fix, codepage is 1252, record is 1 lines);

tabOutput:
Generic LOAD *
Resident tabInput
Where ColNam <>'[group]';

DROP Table tabInput;

hope this helps

regards

Marco