TechEvent 2019: Uses of Row Pattern Matching; Kim Berg Hansen - Trivadis

http://kibeha.dk@kibeha
Uses of Row Pattern Matching
Kim

About me
• Danish geek
• SQL & PL/SQL developer since 2000
• Developer at Trivadis since 2016 http://www.trivadis.com
• Oracle Certified Expert in SQL
• Oracle ACE Director
• SQL quizmaster http://devgym.oracle.com
• Blogger http://kibeha.dk
• Likes to cook and read sci-fi
• Member of Danish Beer Enthusiasts
@kibeha

3 Membership Tiers
• Oracle ACE Director
• Oracle ACE
• Oracle ACE Associate
bit.ly/OracleACEProgram
500+ Technical Experts
Helping Peers Globally
Connect:
Nominate yourself or someone you know: acenomination.oracle.com
@oracleace
Facebook.com/oracleaces
oracle-ace_ww@oracle.com

About Trivadis
• Founded 1994
• 16 locations: Switzerland, Germany, Austria, Denmark and Romania
• 700 specialists
• 260 Service Level Agreements
• Over 4,000 training participants
• Research and development budget:
EUR 5.0 million
• More than 1,900 projects per year at over 800 customers
• Financially self-supporting and sustainably profitable
02-Oct-19 Uses of Row Pattern Matching4

Agenda for Pattern Matching
• Elements in the syntax
• Use cases:
• Stock ticker
• Grouping sequences
• Merge date ranges
• Tablespace growth
• Bin fitting with limited capacity
• Bin fitting in limited number of bins
• Hierarchical child count
• Brief summary

Elements in the syntax

• Example from Data Warehousing Guide chapter on SQL For Pattern Matching
SELECT *
FROM Ticker MATCH_RECOGNIZE (
PARTITION BY symbol
ORDER BY tstamp
MEASURES STRT.tstamp AS start_tstamp,
FINAL LAST(DOWN.tstamp) AS bottom_tstamp,
FINAL LAST(UP.tstamp) AS end_tstamp,
MATCH_NUMBER() AS match_num,
CLASSIFIER() AS var_match
ALL ROWS PER MATCH
AFTER MATCH SKIP TO LAST UP
PATTERN (STRT DOWN+ UP+)
DEFINE
DOWN AS DOWN.price < PREV(DOWN.price),
UP AS UP.price > PREV(UP.price)
) MR
ORDER BY MR.symbol, MR.match_num, MR.tstamp
What‘s it look like

Elements
• PARTITION BY – like analytics split data to work on one partition at a time
• ORDER BY – in which order shall rows be tested whether they match the pattern
• MEASURES – the information we want returned from the match
• ALL ROWS / ONE ROW PER MATCH – return aggregate or detailed info for match
• AFTER MATCH SKIP … – when match found, where to start looking for new match
• PATTERN – regexp like syntax of pattern of defined row classifiers to match
• SUBSET – „union“ a set of classifications into one classification variable
• DEFINE – definition of classification of rows
• FIRST, LAST, PREV, NEXT – navigational functions
• CLASSIFIER(), MATCH_NUMBER() – identification functions

Stock ticker

• Example from Data Warehousing Guide chapter on SQL for Pattern Matching
create table ticker (
symbol varchar2(10)
, day date
, price number
);
insert into ticker values('PLCH', DATE '2011-04-01', 12);
Ticker table

• Look for V shapes = at least one “down” slope followed by at least one “up” slope
select *
from ticker match_recognize (
partition by symbol
order by day
measures strt.day as start_day,
final last(down.day) as bottom_day,
final last(up.day) as end_day,
match_number() as match_num,
classifier() as var_match
all rows per match
after match skip to last up
pattern (strt down+ up+)
define
down as down.price < prev(down.price),
up as up.price > prev(up.price)
) mr
order by mr.symbol, mr.match_num, mr.day;
Stock ticker

• Output of previous slide
SYMBOL DAY START_DAY BOTTOM_DA END_DAY MATCH_NUM VAR_MATCH PRICE
---------- --------- --------- --------- --------- ---------- --------- ----------
PLCH 05-APR-11 05-APR-11 06-APR-11 10-APR-11 1 STRT 25
PLCH 06-APR-11 05-APR-11 06-APR-11 10-APR-11 1 DOWN 12
PLCH 07-APR-11 05-APR-11 06-APR-11 10-APR-11 1 UP 15
Stock ticker

• Previous example ALL ROWS, here ONE ROW per match
select * from ticker match_recognize (
partition by symbol order by day
measures strt.day as start_day,
final last(down.day) as bottom_day,
final last(down.price) as bottom_price,
final last(up.day) as end_day,
match_number() as match_num
one row per match after match skip to last up
pattern (strt down+ up+)
define down as down.price < prev(down.price),
up as up.price > prev(up.price) ) mr
order by mr.symbol, mr.match_num;
SYMBOL START_DAY BOTTOM_DA BOTTOM_PRICE END_DAY MATCH_NUM
---------- --------- --------- ------------ --------- ----------
PLCH 05-APR-11 06-APR-11 12 10-APR-11 1
PLCH 10-APR-11 12-APR-11 15 13-APR-11 2
PLCH 14-APR-11 16-APR-11 12 18-APR-11 3
ONE ROW PER MATCH

• Navigational functions in measure expressions (quiz from devgym.oracle.com)
select symbol, day, price
, up_day, up_avg, up_total
from ticker
match_recognize (
partition by symbol
order by day
measures
final count(up.*) as days_up
, up.price - prev(up.price) as up_day
, (final last(up.price) - strt.price)
/ final count(up.*) as up_avg
, up.price - strt.price as up_total
all rows per match
after match skip to last up
pattern ( strt up+ )
define up as up.price > prev(up.price)
)
order by day;
SYMB DAY PRICE UP_DAY UP_AVG UP_TOTAL
---- --------- ----- ------ ------ --------
PLCH 01-APR-11 12 3.25
PLCH 02-APR-11 17 5 3.25 5
PLCH 03-APR-11 19 2 3.25 7
PLCH 04-APR-11 21 2 3.25 9
PLCH 05-APR-11 25 4 3.25 13
PLCH 06-APR-11 12 3.25
PLCH 07-APR-11 15 3 3.25 3
PLCH 08-APR-11 20 5 3.25 8
PLCH 09-APR-11 24 4 3.25 12
PLCH 10-APR-11 25 1 3.25 13
PLCH 12-APR-11 15 10.00
PLCH 13-APR-11 25 10 10.00 10
PLCH 16-APR-11 12 6.00
PLCH 17-APR-11 14 2 6.00 2
PLCH 18-APR-11 24 10 6.00 12
Measure expressions

Grouping sequences

• https://stewashton.wordpress.com/2014/03/05/12c-match_recognize-grouping-sequences/
• Table of numeric values in some sequential groups
create table ex1 (numval)
as
select 1 from dual union all
select 20 from dual;
Stew Ashton example

• “b” row is a row where numval is exactly one greater than previous rows numval
• Pattern states any row followed by zero or more occurrences of “b” row
select *
from ex1
match_recognize (
order by numval
measures
first(numval) firstval
, last(numval) lastval
, count(*) cnt
pattern (
a b*
)
define
b as numval = prev(numval) + 1
);
FIRSTVAL LASTVAL CNT
---------- ---------- ----------
1 3 3
5 7 3
10 12 3
20 20 1
DEFINE in relation to PREV row

• Analytic method by Aketi Jyuuzou – as efficient, but less self-documenting
select min(numval) firstval
, max(numval) lastval
, count(*) cnt
from (
select numval
, numval - row_number() over (
order by numval
) as grp
from ex1
)
group by grp
order by min(numval);
FIRSTVAL LASTVAL CNT
---------- ---------- ----------
1 3 3
5 7 3
10 12 3
20 20 1
Tabibitosan

Merge date ranges

• https://stewashton.wordpress.com/2015/06/10/merging-overlapping-date-ranges-with-match_recognize/
• Table of date ranges – open-ended end_date (up to but not including)
create table t ( id int, start_date date, end_date date );
insert into t values ( 1, date '2014-01-01', date '2014-01-03');
insert into t values (10, date '2014-02-05', date '2014-02-12');
Date Ranges

• As long as the start date of the next row is smaller than or equal to the highest end date seen
so far, the next row overlaps or adjoins and is merged (replace <= with < for just overlapping)
select *
from t
match_recognize(
order by start_date, end_date
measures
first(start_date) start_date
, max(end_date) end_date
, count(*) c
pattern(
a* b
)
define
a as next(start_date) <= max(end_date)
);
START_DAT END_DATE C
--------- --------- --
01-JAN-14 07-JAN-14 5
23-JAN-14 15-FEB-14 6
Merge overlapping and contiguous ranges

• Add some rows with NULL values
insert into t values (12, null, date '2014-01-01');
insert into t values (13, null, date '2014-01-02');
insert into t values (14, date '2014-02-20', null);
insert into t values (15, date '2014-02-21', null);
NULL for infinity

• Handle null start date as minimum date -4712-01-01
• Handle null end date as maximum date 9999-12-31
select * from t
match_recognize(
order by start_date nulls first
, end_date nulls last
measures
first(start_date) start_date
, nullif(
max(nvl(end_date, date '9999-12-31'))
, date '9999-12-31'
) end_date
, count(*) c
pattern( a* b )
define a as
nvl(next(start_date), date '-4712-01-01')
<= max(nvl(end_date, date '9999-12-31'))
);
START_DAT END_DATE C
--------- --------- --
07-JAN-14 7
23-JAN-14 15-FEB-14 6
19-FEB-14 3
NULL for inifinity

Tablespace growth

• Table storing tablespace size every midnight
create table plch_space (
tabspace varchar2(30)
, sampledate date
, gigabytes number
);
insert into plch_space values ('MYSPACE' , date '2014-02-01', 100);
insert into plch_space values ('YOURSPACE', date '2014-02-06', 50);
insert into plch_space values ('HISSPACE', date '2014-02-06', 100);
From my quizzes on devgym.oracle.com

• FAST defined as 25% growth, SLOW defined as 10-25% growth
• PATTERN states we want to see periods of at least 1 FAST or at least 3 SLOW
select tabspace, spurttype, startdate, startgb, enddate, endgb, avg_daily_gb
from plch_space
match_recognize (
partition by tabspace order by sampledate
measures
classifier() as spurttype
, first(sampledate) as startdate
, first(gigabytes) as startgb
, last(sampledate) as enddate
, next(gigabytes) as endgb
, (next(gigabytes) - first(gigabytes)) / count(*) as avg_daily_gb
one row per match after match skip past last row
pattern ( fast+ | slow{3,} )
define fast as next(gigabytes) / gigabytes >= 1.25
, slow as next(slow.gigabytes) / slow.gigabytes >= 1.10 and
next(slow.gigabytes) / slow.gigabytes < 1.25
)
order by tabspace, startdate;
OR in pattern is |

• Output of the previous slide
TABSPACE SPURTTYPE STARTDATE STARTGB ENDDATE ENDGB AVG_DAILY_GB
------------ ---------- --------- ---------- --------- ---------- ------------
HISSPACE FAST 06-FEB-14 100 06-FEB-14 130 30
HISSPACE FAST 08-FEB-14 145 08-FEB-14 200 55
HISSPACE SLOW 09-FEB-14 200 12-FEB-14 315 28.75
MYSPACE SLOW 02-FEB-14 103 05-FEB-14 160 14.25
MYSPACE FAST 07-FEB-14 165 07-FEB-14 210 45
YOURSPACE FAST 07-FEB-14 53 08-FEB-14 97 22
Growth alert report

select tabspace, spurttype, startdate
, min(gigabytes) keep (dense_rank first order by sampledate) startgb
, max(sampledate) enddate
, max(nextgb) keep (dense_rank last order by sampledate) endgb
, avg(daily_gb) avg_daily_gb
from (
select tabspace, spurttype, sampledate, gigabytes, nextgb, daily_gb
, last_value(spurtstartdate ignore nulls) over (
partition by tabspace, spurttype order by sampledate
rows between unbounded preceding and current row
) startdate
from (
select tabspace, spurttype, sampledate, gigabytes, nextgb, daily_gb
, case
when spurttype is not null and
( lag(spurttype) over (
) is null
or
lag(spurttype) over (
) != spurttype
)
...
Analytic alternative

...
then sampledate
end spurtstartdate
from (
select tabspace, sampledate, gigabytes, nextgb, nextgb - gigabytes daily_gb
, case
when nextgb >= gigabytes * 1.25 then 'FAST'
when nextgb >= gigabytes * 1.10 then 'SLOW'
end spurttype
from (
select tabspace, sampledate, gigabytes
, lead(gigabytes) over (
) nextgb
from plch_space
) ) )
where spurttype is not null
)
group by tabspace, spurttype, startdate
having count(*) >= case spurttype
when 'FAST' then 1
when 'SLOW' then 3
end
order by tabspace, startdate;
Analytic alternative (continued)

Bin fitting –
limited capacity

• https://stewashton.wordpress.com/2014/03/03/database-12c-match_recognize-for-all-sizes-of-data/
• Create groups of consecutive study_site with sum(cnt) at most 65.000
create table t (
study_site number
, cnt number
);
insert into t (study_site,cnt) values (1001,3407);
Stew Ashton example

• Aggregate SUM in Define is "running“ semantic
• Pattern "a+" continues matching while rolling sum(cnt) <= 65.000
select * from t
match_recognize (
order by study_site
measures
first(study_site) first_site
, last(study_site) last_site
, sum(cnt) sum_cnt
one row per match
after match skip past last row
pattern ( a+ )
define
a as sum(cnt) <= 65000
);
FIRST_SITE LAST_SITE SUM_CNT
---------- ---------- ----------
1001 1022 48081
1023 1044 62203
1045 1045 3360
Match until rolling sum reaches limit

• Previous slide was criteria had to order by STUDY_SITE
• Ordering by CNT descending can "pack" the data a bit better
select * from t
match_recognize (
order by cnt desc, study_site
measures
count(*) sites
, sum(cnt) sum_cnt
, min(cnt) min_cnt
, max(cnt) max_cnt
one row per match
after match skip past last row
pattern ( a+ )
define
a as sum(cnt) <= 65000
);
SITES SUM_CNT MIN_CNT MAX_CNT
------ -------- -------- --------
6 62588 6005 25865
22 51056 3 5282

• Better (yet simple) "best fit" approximation by interleaved ordering of large/small
• Largest, smallest, second-largest, second-smallest, third-largest, third-smallest, etc.
select * from (
select study_site, cnt
, least(
row_number() over (
order by cnt
)
, row_number() over (
order by cnt desc
)
) rn
from t
)
match_recognize (
order by rn, cnt desc, study_site
...
SITES SUM_CNT MIN_CNT MAX_CNT
------ -------- -------- --------
11 64154 3 25865
17 49490 885 5282

Bin fitting –
limited number of bins

• https://stewashton.wordpress.com/2014/06/06/bin-fitting-problems-with-sql/
• We want to fill 3 bins so each bin sum(item_value) is as near equal as possible
create table items
as
select level item_name, level
item_value
from dual
connect by level <= 10;
select *
from items
order by item_name;
ITEM_NAME ITEM_VALUE
---------- ----------
1 1
2 2
3 3
4 4
5 5
6 6
7 7
8 8
9 9
10 10
Stew Ashton example

• First, order the items by value in descending order
• Then, assign each item to whatever bin has the smallest sum so far
select * from items
match_recognize (
order by item_value desc
measures
to_number(substr(classifier(),4)) bin#,
sum(bin1.item_value) bin1,
sum(bin2.item_value) bin2,
sum(bin3.item_value) bin3
all rows per match
pattern ( (bin1|bin2|bin3)* )
define
bin1 as count(bin1.*) = 1
or sum(bin1.item_value)-bin1.item_value
<= least(sum(bin2.item_value), sum(bin3.item_value))
, bin2 as count(bin2.*) = 1
or sum(bin2.item_value)-bin2.item_value
<= sum(bin3.item_value)
);
Fill 3 bins equally

ITEM_VALUE BIN# BIN1 BIN2 BIN3 ITEM_NAME
---------- ---------- ---------- ---------- ---------- ----------
10 1 10 10
9 2 10 9 9
8 3 10 9 8 8
7 3 10 9 15 7
6 2 10 15 15 6
5 1 15 15 15 5
4 1 19 15 15 4
3 2 19 18 15 3
2 3 19 18 17 2
1 3 19 18 18 1
Almost equally filled

Hierarchical child count

• http://www.kibeha.dk/2015/07/row-pattern-matching-nested-within.html
• CONNECT BY in scalar subquery
select empno
, lpad(' ', (level-1)*2) || ename as ename
, (
select count(*)
from emp sub
start with sub.mgr = emp.empno
connect by sub.mgr = prior sub.empno
) subs
from emp
start with mgr is null
connect by mgr = prior empno
order siblings by empno;
EMPNO ENAME SUBS
----- ------------ -----
7839 KING 13
7566 JONES 4
7788 SCOTT 1
7876 ADAMS 0
7902 FORD 1
7369 SMITH 0
7698 BLAKE 5
7499 ALLEN 0
7521 WARD 0
7654 MARTIN 0
7844 TURNER 0
7900 JAMES 0
7782 CLARK 1
7934 MILLER 0
How many subordinates for each employee

• Using AFTER MATCH SKIP TO NEXT ROW allows “nesting” of matches
• Identical output as previous slide
with hierarchy as (
select lvl, empno, ename, rownum as rn
from (
select level as lvl, empno, ename
from emp
order siblings by empno
)
)
select empno
, lpad(' ', (lvl-1)*2) || ename as ename
, subs
from hierarchy
...
...
match_recognize (
order by rn
measures
strt.rn as rn
, strt.lvl as lvl
, strt.empno as empno
, strt.ename as ename
, count(higher.lvl) as subs
one row per match
after match skip to next row
pattern ( strt higher* )
define higher as
higher.lvl > strt.lvl
)
order by rn;
Pattern matching instead of scalar subquery

• See details of what is happening with ALL ROWS PER MATCH
with hierarchy as (
from (
from emp
) )
select mn, rn, empno
, roll, subs, cls
, stno, stname, hino, hiname
from hierarchy
match_recognize (
order by rn
...
...
measures
match_number() as mn
, classifier() as cls
, strt.empno as stno
, strt.ename as stname
, higher.empno as hino
, higher.ename as hiname
, count(higher.lvl) as roll
, final count(higher.lvl) as subs
all rows per match
define higher as
)
order by mn, rn;
ALL ROWS PER MATCH

MN RN EMPNO ENAME ROLL SUBS CLS STNO STNAME HINO HINAME
--- --- ----- ------------ ---- ---- ------ ----- ------ ----- ------
1 1 7839 KING 0 13 STRT 7839 KING
1 2 7566 JONES 1 13 HIGHER 7839 KING 7566 JONES
1 3 7788 SCOTT 2 13 HIGHER 7839 KING 7788 SCOTT
1 4 7876 ADAMS 3 13 HIGHER 7839 KING 7876 ADAMS
1 5 7902 FORD 4 13 HIGHER 7839 KING 7902 FORD
1 6 7369 SMITH 5 13 HIGHER 7839 KING 7369 SMITH
1 7 7698 BLAKE 6 13 HIGHER 7839 KING 7698 BLAKE
1 8 7499 ALLEN 7 13 HIGHER 7839 KING 7499 ALLEN
1 9 7521 WARD 8 13 HIGHER 7839 KING 7521 WARD
1 10 7654 MARTIN 9 13 HIGHER 7839 KING 7654 MARTIN
1 11 7844 TURNER 10 13 HIGHER 7839 KING 7844 TURNER
1 12 7900 JAMES 11 13 HIGHER 7839 KING 7900 JAMES
1 13 7782 CLARK 12 13 HIGHER 7839 KING 7782 CLARK
1 14 7934 MILLER 13 13 HIGHER 7839 KING 7934 MILLER
2 2 7566 JONES 0 4 STRT 7566 JONES
2 3 7788 SCOTT 1 4 HIGHER 7566 JONES 7788 SCOTT
2 4 7876 ADAMS 2 4 HIGHER 7566 JONES 7876 ADAMS
2 5 7902 FORD 3 4 HIGHER 7566 JONES 7902 FORD
2 6 7369 SMITH 4 4 HIGHER 7566 JONES 7369 SMITH
...
ALL ROWS PER MATCH

• PIVOT just to visualize the output which rows are part of what match
with hierarchy as (
from (
from emp
) )
select rn, empno, ename
, case "1" when 1 then 'XX' end "1"
...
...
...
from (
select mn, rn, empno
, lpad(' ', (lvl-1)*2) || ename as
ename
from hierarchy
match_recognize (
order by rn
measures match_number() as mn
all rows per match
define higher as higher.lvl > strt.lvl
))
pivot (
count(*)
for mn in (1,2,3,4,5,6,7,8,9,10,11,12,13,14)
) order by rn;
PIVOT

• Output of the previous slide
RN EMPNO ENAME 1 2 3 4 5 6 7 8 9 10 11 12 13 14
--- ----- ------------ -- -- -- -- -- -- -- -- -- -- -- -- -- --
1 7839 KING XX
2 7566 JONES XX XX
3 7788 SCOTT XX XX XX
4 7876 ADAMS XX XX XX XX
5 7902 FORD XX XX XX
6 7369 SMITH XX XX XX XX
7 7698 BLAKE XX XX
8 7499 ALLEN XX XX XX
9 7521 WARD XX XX XX
10 7654 MARTIN XX XX XX
11 7844 TURNER XX XX XX
12 7900 JAMES XX XX XX
13 7782 CLARK XX XX
14 7934 MILLER XX XX XX
PIVOT

• Could wrap entire thing in inline view and filter on “subs > 0”
• But much simpler just to change * into +
with hierarchy as (
from (
from emp
)
)
select empno
, subs
from hierarchy
...
...
match_recognize (
order by rn
measures
strt.rn as rn
, strt.lvl as lvl
, strt.empno as empno
, strt.ename as ename
, count(higher.lvl) as subs
one row per match
pattern ( strt higher+ )
define higher as
)
order by rn;
Only those with subordinates?

EMPNO ENAME SUBS
----- ------------ ----
7839 KING 13
7566 JONES 4
7788 SCOTT 1
7902 FORD 1
7698 BLAKE 5
7782 CLARK 1
Only those with subordinates!

• Create BIGEMP table with emp LARRY on top of pyramid of 14.001 employees
create table bigemp as
select 1 empno
, 'LARRY' ename
, cast(null as number) mgr
from dual
union all
select dum.dum * 10000 + empno empno
, ename || '#' || dum.dum ename
, coalesce(dum.dum * 10000 + mgr, 1) mgr
from emp
cross join (
select level dum
from dual
connect by level <= 1000
) dum;
Scalability

• Scalar subquery with CONNECT BY on left 30x slower, 8455x more gets, 9252x more sorts
than MATCH_RECOGNIZE method on right
14001 rows selected.
Elapsed: 00:00:11.61
Statistics
--------------------------------------------
0 recursive calls
0 db block gets
465005 consistent gets
0 physical reads
0 redo size
435280 bytes sent via SQL*Net to client
10763 bytes received via SQL*Net from...
935 SQL*Net roundtrips to/from client
37008 sorts (memory)
0 sorts (disk)
14001 rows processed
14001 rows selected.
Elapsed: 00:00:00.35
Statistics
--------------------------------------------
1 recursive calls
0 db block gets
55 consistent gets
0 physical reads
0 redo size
435280 bytes sent via SQL*Net to client
10763 bytes received via SQL*Net from...
935 SQL*Net roundtrips to/from client
4 sorts (memory)
0 sorts (disk)
14001 rows processed
Scalability

Brief summary

MATCH_RECOGNIZE - A “swiss army knife” tool
• Brilliant when applied “BI style” like stock ticker analysis examples
• But applicable to many other cases too
• When you have some problem crossing row boundaries and feel you have to “stretch” even
the capabilities of analytics, try a pattern based approach:
• Rephrase (in natural language) your requirements in terms of what classifies the rows
you are looking for
• Turn that into pattern matching syntax classifying individual rows in DEFINE and how the
classified rows should appear in PATTERN
• As with analytics, it might feel daunting at first, but once you start using pattern matching, it
will become just another tool in your SQL toolbox

http://kibeha.dk@kibeha
Questions & Answers
This presentation http://bit.ly/kibeha_patmatch4_pptx
Script with all the code http://bit.ly/kibeha_patmatch4_sql
Webinar http://bit.ly/patternmatch
Webinar scripts http://bit.ly/patternmatchsamples
Stew Ashton https://stewashton.wordpress.com/category/match_recognize/
kim.berghansen@trivadis.com

TechEvent 2019: Uses of Row Pattern Matching; Kim Berg Hansen - Trivadis

More Related Content

What's hot (20)

Similar to TechEvent 2019: Uses of Row Pattern Matching; Kim Berg Hansen - Trivadis (20)

More from Trivadis (20)

Recently uploaded (20)

TechEvent 2019: Uses of Row Pattern Matching; Kim Berg Hansen - Trivadis