Tabibitosan method tutorial by Aketi Jyuuzou

Aketi Jyuuzou

This thread explains Tabibitosan method B-)

*****************************************************************
1. What is Tabibitosan ? ?:|

"Tabibitosan" is Japanese language.
"Tabibitosan" is one of math problem.

http://www.manabinoba.com/index.cfm/7,757,33,1,html
http://yslibrary.cool.ne.jp/sansub1801.html

I do not know what "Tabibitosan" is called in English.
If you know, please teach me.

In Japanese -> English dictionary,
"Tabibito" means "pilgrim","viator" or "wayfaring man".

Japanese "Tabibitosan" book
http://www.amazon.co.jp/dp/4895243702

*****************************************************************
2. What is basic usage of "Tabibitosan method" ? ?:|

"Tabibitosan method" is very effective when we want to collect Rows which are continued a lot of condition.

**************************
For exsample1 (collect Rows continued Number)

create table Ex1 (NumVal primary key) as
select  1 from dual union
select  2 from dual union
select  3 from dual union
select  5 from dual union
select  6 from dual union
select  7 from dual union
select 10 from dual union
select 11 from dual union
select 12 from dual union
select 20 from dual union
select 21 from dual;

I will derive resultSet which collects Rows continued NumVal.

select min(NumVal),max(NumVal),count(*)
from (select NumVal,
      NumVal-Row_Number() over(order by NumVal)
      as disTance
      from Ex1)
group by disTance
order by min(NumVal);

MIN(NUMVAL) MAX(NUMVAL)   COUNT(*)
----------- ----------- ----------
          1           3          3
          5           7          3
         10          12          3
         20          21          2

In above solution,
I imagined that there are 2 people called X and A.
X always walks 1m. (Row_Number() over(order by NumVal))
A always walks more than 1m. (NumVal)

Then,I derived distance between X and A.
Then, I made group using these distance. (group by disTance)

**************************
For exsample2 (collect Rows continued Days)

create table Ex2 (DateVal primary key) as
select date '2009-12-10' from dual union
select date '2009-12-11' from dual union
select date '2009-12-12' from dual union
select date '2009-12-16' from dual union
select date '2009-12-17' from dual union
select date '2009-12-20' from dual;

select min(DateVal),max(DateVal),count(*)
from (select DateVal,
      DateVal-Row_Number() over(order by DateVal)
      as disTance
      from Ex2)
group by disTance
order by min(DateVal);

MIN(DATE MAX(DATE   COUNT(*)
-------- -------- ----------
09-12-10 09-12-12          3
09-12-16 09-12-17          2
09-12-20 09-12-20          1

Aketi Jyuuzou

**************************
For exsample3 (collect Rows continued Months)

create table Ex3 (DateVal primary key) as
select date '2009-09-01' from dual union
select date '2009-10-01' from dual union
select date '2009-12-01' from dual union
select date '2010-01-01' from dual union
select date '2010-02-01' from dual union
select date '2010-04-01' from dual;

select min(DateVal),max(DateVal),count(*)
from (select DateVal,
       extract(year  from DateVal)*12
      +extract(month from DateVal)
      -Row_Number() over(order by DateVal)
      as disTance
      from Ex3)
group by disTance
order by min(DateVal);

MIN(DATE MAX(DATE   COUNT(*)
-------- -------- ----------
09-09-01 09-10-01          2
09-12-01 10-02-01          3
10-04-01 10-04-01          1

*****************************************************************
3. What is complex usage of "Tabibitosan method" ? ?:|

**************************
For exsample4 (collect Rows continued and ID is equal)
from 450745

create table Ex4 (ID,Val,SortKey) as
select 1, 5, 1 from dual union all
select 1,10, 2 from dual union all
select 2, 2, 3 from dual union all
select 2, 5, 4 from dual union all
select 1,15, 5 from dual union all
select 3,25, 6 from dual union all
select 3,10, 7 from dual union all
select 3, 5, 8 from dual union all
select 3,15, 9 from dual union all
select 4, 5,10 from dual;

select ID,min(Val),max(Val),count(*)
from (select ID,Val,SortKey,
       Row_Number() over(order by SortKey)
      -Row_Number() over(partition by ID order by SortKey)
      as disTance
      from Ex4)
group by ID,disTance
order by min(SortKey);

ID MIN(VAL) MAX(VAL) COUNT(*)
-- -------- -------- --------
 1        5       10        2
 2        2        5        2
 1       15       15        1
 3        5       25        4
 4        5        5        1

In above solution,
I imagined that there are 5 people called X,A,B,C, and D.

X always walks 1m. (Row_Number() over(order by SortKey))
When ID = 1, A walks 1m. (Row_Number() over(partition by ID order by SortKey))
When ID = 2, B walks 1m. (Row_Number() over(partition by ID order by SortKey))
When ID = 3, C walks 1m. (Row_Number() over(partition by ID order by SortKey))
When ID = 4, D walks 1m. (Row_Number() over(partition by ID order by SortKey))

Then,
I derived distance between X and A.
I derived distance between X and B.
I derived distance between X and C.
I derived distance between X and D.

Then, I made group using these distances. (group by ID,disTance)

**************************
For exsample5 (collect Rows continued and Val1 is equal and Val2 is equal)
from 941878

create table mytable (sortKey,Val1,Val2) as
select 1,'A','X' from dual union all
select 2,'A','X' from dual union all
select 3,'B','Y' from dual union all
select 4,'B','Y' from dual union all
select 5,'A','X' from dual union all
select 5,'B','X' from dual union all
select 6,'A','Y' from dual union all
select 7,'B','Y' from dual union all
select 7,'A','Y' from dual union all
select 8,'A','Y' from dual;

select Val1,Val2,min(sortKey) as sta,max(sortKey) as end
from (select sortKey,Val1,Val2,
       dense_rank() over(order by sortKey)
      -Row_Number() over(partition by Val1,Val2
                         order by sortKey)
      as distance
       from mytable)
group by Val1,Val2,distance
order by min(sortKey);

Val1 Val2 sta end
---- ---- --- ---
A    X      1   2
B    Y      3   4
B    X      5   5
A    X      5   5
A    Y      6   8
B    Y      7   7

In above solution,
I imagined that there are 5 people called X,A,B,C, and D.

X always walks 1m. (dense_rank() over(order by sortKey))
When Val1 = 'A' and Val2 = 'X', A walks 1m. (Row_Number() over(partition by Val1,Val2 order by SortKey))
When Val1 = 'A' and Val2 = 'Y', B walks 1m. (Row_Number() over(partition by Val1,Val2 order by SortKey))
When Val1 = 'B' and Val2 = 'X', C walks 1m. (Row_Number() over(partition by Val1,Val2 order by SortKey))
When Val1 = 'B' and Val2 = 'Y', D walks 1m. (Row_Number() over(partition by Val1,Val2 order by SortKey))

Then,
I derived distance between X and A.
I derived distance between X and B.
I derived distance between X and C.
I derived distance between X and D.

Then, I made group using these distances. (group by Val1,Val2,disTance)

*****************************************************************
4. What threads are using "Tabibitosan method" ?:|

I will write that threads sorted by easy.

basic usage
1005633
866707

complex usage
920572
450745
1028579
941878

*****************************************************************
5. What URLS mention "Tabibitosan method" ?:|

Boneist's Oracle blog
http://boneist-oracle.livejournal.com/7389.html

Alex Nuijten's Presentation and that PDF document.
http://caat.odtug.com/odtug_caat/caat_abstract_reports.display_presenter_abstract?conference_id=81&presenter_id=117&abstract_id=186

*****************************************************************
OracleSQLPuzzle (MyHomePage)
http://www.geocities.jp/oraclesqlpuzzle/

sister thread
Boolean algebra in some Rows tutorial by Aketi Jyuuzou 1007678

Karthick2003

Thanks, Learned something new today. This kind of problem comes up quite often.

Keep up the good work. Much Appreciated!!

743014

here is another example using this with dates. It groups ids by consecutive weeks

WITH A AS
( SELECT 1 ID,5 Val, to_date('2009-08-24','yyyy-mm-dd') StartWeek, to_date('2009-08-30','yyyy-mm-dd') EndWeek FROM dual UNION ALL 
  SELECT 1,10, to_date('2009-08-31','yyyy-mm-dd'), to_date('2009-09-06','yyyy-mm-dd') FROM dual UNION ALL
  SELECT 1,2, to_date('2009-09-07','yyyy-mm-dd'), to_date('2009-09-13','yyyy-mm-dd') FROM dual UNION ALL  
  SELECT 2,2, to_date('2009-09-07','yyyy-mm-dd'), to_date('2009-09-13','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 2,5, to_date('2009-09-14','yyyy-mm-dd'), to_date('2009-09-20','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 3,5, to_date('2009-09-14','yyyy-mm-dd'), to_date('2009-09-20','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 1,15, to_date('2009-09-21','yyyy-mm-dd'), to_date('2009-09-27','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 3,25, to_date('2009-09-28','yyyy-mm-dd'), to_date('2009-10-04','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 3,10, to_date('2009-10-05','yyyy-mm-dd'), to_date('2009-10-11','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 3,5, to_date('2009-10-12','yyyy-mm-dd'), to_date('2009-10-18','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 3,15, to_date('2009-10-19','yyyy-mm-dd'), to_date('2009-10-25','yyyy-mm-dd') FROM dual UNION ALL 
  SELECT 4,5, to_date('2009-10-26','yyyy-mm-dd'), to_date('2009-11-01','yyyy-mm-dd') FROM dual)
,
B AS
( SELECT ID,        Val,        StartWeek,        EndWeek,
         (Row_Number() over(ORDER BY StartWeek) - Row_Number() over(partition BY ID ORDER BY StartWeek)) AS makeGroup
 FROM A)
SELECT ID,       SUM(Val) AS SumVal,       MIN(StartWeek),       MAX(EndWeek)
FROM B
GROUP BY ID,makeGroup
ORDER BY MIN(StartWeek);

ID            SumVAL        MIN(STARTWEEK)      MAX(ENDWEEK)
1             17            8/24/2009           9/13/2009
2             7             9/7/2009            9/20/2009
3             5             9/14/2009	        9/20/2009
1             15            9/21/2009	        9/27/2009
3             55            9/28/2009	        10/25/2009
4             5             10/26/2009          11/1/2009

Frank Kulash

Hi,

Aketi Jyuuzou wrote:
...
1. What is Tabibitosan ? ?:|

"Tabibitosan" is Japanese language.
"Tabibitosan" is one of math problem.

What is the math problem? It sounds interesting.

http://www.manabinoba.com/index.cfm/4,757,73,html?year=2002
http://yslibrary.cool.ne.jp/sansub1801.html

I love the graphics on the second link!

AlanWms

Okay, I am a little dense, I was going to ask for an explanation, in very simple terms, then I figured it out. Basically it is subtracting (or adding if your list is descending) the row number from an ordered list of numbers (or dates) to find consecutive ranges of numbers (or dates). Pretty slick.

As an example, lets take some integers with gaps in the sequence:

Num      Row    Distance (Num-Row)
===      ===    =============
2        1        1
3        2        1
4        3        1
5        4        1
6        5        1
7        6        1
12       7        5
13       8        5
14       9        5 
15       10       5
16       11       5
17       12       5
20       13       7
21       14       7
etc ...

Note that the "Distance" delineates consecutive groups, you then group by this etc. Maybe everyone else who viewed this thread got it immediately...

Thanks Aketi, this is a neat trick!

Solomon Yakobson

AlanWms wrote:
Okay, I am a little dense, I was going to ask for an explanation, in very simple terms, then I figured it out.

It looks like one of the simplest cases of grouping "loosely grouped" data.

SY.

730428

Hi, here is another way to solve example 1 using hierarchical queries:

SQL> select min(CONNECT_BY_ROOT NUMVAL) minval, NUMVAL maxval, max(NUMVAL-CONNECT_BY_ROOT NUMVAL+1)  countval
  2  from ex1
  3  where connect_by_isleaf=1
  4  connect by numval=prior numval+1
  5  group by numval
  6  order by 1;

    MINVAL     MAXVAL   COUNTVAL
---------- ---------- ----------
         1          3          3
         5          7          3
        10         12          3
        20         21          2

The other examples can be easily derived...

Max
[My Italian Oracle blog|http://oracleitalia.wordpress.com/2009/12/27/inviare-email-dal-db-utilizzando-utl_smtp/]

Aketi Jyuuzou

@Massimo Ruocchio

Nice one :-)
Hehe I have arranged your solution B-)

select min(CONNECT_BY_ROOT NUMVAL) minval,
NUMVAL maxval, max(Level) countval
from ex1
where connect_by_isleaf=1
connect by numval=prior numval+1
group by numval
order by minval;

MINVAL  MAXVAL  COUNTVAL
------  ------  --------
     1       3         3
     5       7         3
    10      12         3
    20      21         2

Aketi Jyuuzou

5. What is rival solution of Tabibitosan method ? ?:|

There is the rival solution.
Rival solution derives willSum using Lag function and case expression
And Then derive runSum of willSum using sum function.

select min(NumVal),max(NumVal),count(*)
from (select NumVal,
      sum(willSum) over(order by NumVal) as GID
      from (select NumVal,
            case when NumVal-1
               = Lag(NumVal) over(order by NumVal)
                 then 0 else 1 end as willSum
            from Ex1))
group by GID
order by GID;

MIN(NUMVAL)  MAX(NUMVAL)   COUNT(*)
-----------  -----------  ---------
          1            3          3
          5            7          3
         10           12          3
         20           21          2

But This rival solution needs 2 InLineView.
Altough TabibitoSan method needs only 1 InLineView.

Therefore if we can use TabibitoSan method,we should use TabibitoSan method.

for example
In below cases,we cannot use TabibitoSan method.
969005
956106

Etbin

Hats off. A clear and concise treatment of a non elementary group of problems. I like the pilgrim interpretation. Never thought about using it when I was asked to explain a solution of this kind. Tabibitosan - a word to remember.

Regards

Etbin