Обсуждение: transposed query?
Hi;
I've got problems with a query. I'm not sure if it is possible to do this
with Postgres, although I think it should be.
I had resolved these kind of queryes in MSAccess, where they are called
"cross references querys" but I can't find my way in Postgres.
My table 'muestras_rambla' is like:
date | id_punto | muestra | flow | n_nitrato ...
--------------------------+----------+---------+---------+-----------
06/02/2003 00:00:00 CET | 1 | 1 | 699.462 | 18.44
06/02/2003 00:00:00 CET | 1 | 2 | 699.462 | 13.79
06/02/2003 00:00:00 CET | 2 | 1 | 341.05 | 11.15
06/02/2003 00:00:00 CET | 2 | 2 | 341.05 | 17.3
06/02/2003 00:00:00 CET | 3 | 1 | 514.05 | 22.17
06/02/2003 00:00:00 CET | 3 | 2 | 514.05 | 16.62
06/02/2003 00:00:00 CET | 4 | 1 | 466.884 | 18.99
06/02/2003 00:00:00 CET | 4 | 2 | 466.884 | 19.02
06/02/2003 00:00:00 CET | 6 | 1 | 30.012 | 5.91
06/02/2003 00:00:00 CET | 6 | 2 | 30.012 | 4.15
19/02/2003 00:00:00 CET | 1 | 1 | 911.43 | 17.66
19/02/2003 00:00:00 CET | 1 | 2 | 911.43 | 9.23
...
where I 'id_punto' can take the values from 1 to 6, and muestra is 1 or 2.
Commonly, for each 'date' and 'id_punto' we have two samples (muestra), and
the 'flow' is also commonly the same for the two samples.
The case is that for every 'date' I want to have the 'time series' of flows
in the way:
date 1 2 3 4 6
03/10/2002 606.92 323.08 440.69 393.61 5.49
17/10/2002 348.19 400.32 319.33 211.26 2.53
29/10/2002 411.57 409.66 310.37 362.20 9.36
14/11/2002 446.96 373.72 302.79 348.69 5.18
26/11/2002 381.20 386.85 307.77 240.75 6.41
12/12/2002 416.84 307.03 317.94 249.36 5.64
26/12/2002 743.04 380.30 539.94 521.60 3.69
09/01/2003 403.07 301.16 370.30 379.41 3.69
...
where the names of the rows are the values in 'id_punto' and the values of
the data are the average (for every 'id_punto' and 'date') between the two
samples (muestra).
In MSaccess this is solve with:
TRANSFORM avg(muestras_rambla.flow) AS avg_flow
SELECT muestras_rambla.date
FROM muestras_rambla
GROUP BY muestras_rambla.date
PIVOT muestras_rambla.id_punto;
All I can do in postgres is :
--------
SELECT muestras_rambla.fecha, id_punto, avg(caudal) AS avg_caudal
FROM muestras_rambla
GROUP BY fecha, id_punto
--------
to get:
fecha | id_punto | avg_caudal
--------------------------+----------+------------------
03/10/2002 00:00:00 CEST | 1 | 606.924
03/10/2002 00:00:00 CEST | 2 | 323.079
03/10/2002 00:00:00 CEST | 3 | 440.690
...
But, how could manage this to get the 'time series' table?
Thanks for your help and best wishes
Javier
Nick Barr wrote: > javier garcia - CEBAS wrote: > >> Hi; >> I've got problems with a query. I'm not sure if it is possible to do >> this with Postgres, although I think it should be. >> >> I had resolved these kind of queryes in MSAccess, where they are >> called "cross references querys" but I can't find my way in Postgres. >> >> My table 'muestras_rambla' is like: >> date | id_punto | muestra | flow | n_nitrato ... >> --------------------------+----------+---------+---------+----------- >> 06/02/2003 00:00:00 CET | 1 | 1 | 699.462 | 18.44 >> 06/02/2003 00:00:00 CET | 1 | 2 | 699.462 | 13.79 >> 06/02/2003 00:00:00 CET | 2 | 1 | 341.05 | 11.15 >> 06/02/2003 00:00:00 CET | 2 | 2 | 341.05 | 17.3 >> 06/02/2003 00:00:00 CET | 3 | 1 | 514.05 | 22.17 >> 06/02/2003 00:00:00 CET | 3 | 2 | 514.05 | 16.62 >> 06/02/2003 00:00:00 CET | 4 | 1 | 466.884 | 18.99 >> 06/02/2003 00:00:00 CET | 4 | 2 | 466.884 | 19.02 >> 06/02/2003 00:00:00 CET | 6 | 1 | 30.012 | 5.91 >> 06/02/2003 00:00:00 CET | 6 | 2 | 30.012 | 4.15 >> 19/02/2003 00:00:00 CET | 1 | 1 | 911.43 | 17.66 >> 19/02/2003 00:00:00 CET | 1 | 2 | 911.43 | 9.23 >> ... >> >> where I 'id_punto' can take the values from 1 to 6, and muestra is 1 >> or 2. Commonly, for each 'date' and 'id_punto' we have two samples >> (muestra), and the 'flow' is also commonly the same for the two samples. >> >> The case is that for every 'date' I want to have the 'time series' of >> flows in the way: >> date 1 2 3 4 6 >> 03/10/2002 606.92 323.08 440.69 393.61 5.49 >> 17/10/2002 348.19 400.32 319.33 211.26 2.53 >> 29/10/2002 411.57 409.66 310.37 362.20 9.36 >> 14/11/2002 446.96 373.72 302.79 348.69 5.18 >> 26/11/2002 381.20 386.85 307.77 240.75 6.41 >> 12/12/2002 416.84 307.03 317.94 249.36 5.64 >> 26/12/2002 743.04 380.30 539.94 521.60 3.69 >> 09/01/2003 403.07 301.16 370.30 379.41 3.69 >> ... >> >> where the names of the rows are the values in 'id_punto' and the >> values of the data are the average (for every 'id_punto' and 'date') >> between the two samples (muestra). >> >> In MSaccess this is solve with: >> TRANSFORM avg(muestras_rambla.flow) AS avg_flow >> SELECT muestras_rambla.date >> FROM muestras_rambla >> GROUP BY muestras_rambla.date >> PIVOT muestras_rambla.id_punto; >> >> All I can do in postgres is : >> -------- >> SELECT muestras_rambla.fecha, id_punto, avg(caudal) AS avg_caudal >> FROM muestras_rambla >> GROUP BY fecha, id_punto >> -------- >> to get: >> fecha | id_punto | avg_caudal >> --------------------------+----------+------------------ >> 03/10/2002 00:00:00 CEST | 1 | 606.924 >> 03/10/2002 00:00:00 CEST | 2 | 323.079 >> 03/10/2002 00:00:00 CEST | 3 | 440.690 >> ... >> But, how could manage this to get the 'time series' table? >> >> Thanks for your help and best wishes >> >> Javier >> >> ---------------------------(end of broadcast)--------------------------- >> TIP 5: Have you checked our extensive FAQ? >> >> http://www.postgresql.org/docs/faqs/FAQ.html >> > > How about: > > > SELECT > date, > (SELECT s1.flow FROM muestras_rambla s1 WHERE s1.date=t1.date AND > s1.id_punto=1) AS flow_1, > (SELECT s2.flow FROM muestras_rambla s2 WHERE s2.date=t1.date AND > s1.id_punto=2) AS flow_2, > (SELECT s3.flow FROM muestras_rambla s3 WHERE s3.date=t1.date AND > s1.id_punto=3) AS flow_3, > (SELECT s4.flow FROM muestras_rambla s4 WHERE s4.date=t1.date AND > s1.id_punto=4) AS flow_4, > (SELECT s6.flow FROM muestras_rambla s6 WHERE s6.date=t1.date AND > s1.id_punto=6) AS flow_6 > FROM > muestras_rambla t1 > GROUP BY > t1.date; > > > This uses sub-queries in the Select clause to get the data in columns. > There may be another way. > > > > Nick Barr > Ooops, having tested that query I found a typo (or four). Try: SELECT date, (SELECT s1.flow FROM muestras_rambla s1 WHERE s1.date=t1.date AND s1.id_punto=1) AS flow_1, (SELECT s2.flow FROM muestras_rambla s2 WHERE s2.date=t1.date AND s2.id_punto=2) AS flow_2, (SELECT s3.flow FROM muestras_rambla s3 WHERE s3.date=t1.date AND s3.id_punto=3) AS flow_3, (SELECT s4.flow FROM muestras_rambla s4 WHERE s4.date=t1.date AND s4.id_punto=4) AS flow_4, (SELECT s6.flow FROM muestras_rambla s6 WHERE s6.date=t1.date AND s6.id_punto=6) AS flow_6 FROM muestras_rambla t1 GROUP BY t1.date; Nick Barr
javier garcia - CEBAS wrote: > Hi; > I've got problems with a query. I'm not sure if it is possible to do this > with Postgres, although I think it should be. > > I had resolved these kind of queryes in MSAccess, where they are called > "cross references querys" but I can't find my way in Postgres. > > My table 'muestras_rambla' is like: > date | id_punto | muestra | flow | n_nitrato ... > --------------------------+----------+---------+---------+----------- > 06/02/2003 00:00:00 CET | 1 | 1 | 699.462 | 18.44 > 06/02/2003 00:00:00 CET | 1 | 2 | 699.462 | 13.79 > 06/02/2003 00:00:00 CET | 2 | 1 | 341.05 | 11.15 > 06/02/2003 00:00:00 CET | 2 | 2 | 341.05 | 17.3 > 06/02/2003 00:00:00 CET | 3 | 1 | 514.05 | 22.17 > 06/02/2003 00:00:00 CET | 3 | 2 | 514.05 | 16.62 > 06/02/2003 00:00:00 CET | 4 | 1 | 466.884 | 18.99 > 06/02/2003 00:00:00 CET | 4 | 2 | 466.884 | 19.02 > 06/02/2003 00:00:00 CET | 6 | 1 | 30.012 | 5.91 > 06/02/2003 00:00:00 CET | 6 | 2 | 30.012 | 4.15 > 19/02/2003 00:00:00 CET | 1 | 1 | 911.43 | 17.66 > 19/02/2003 00:00:00 CET | 1 | 2 | 911.43 | 9.23 > ... > > where I 'id_punto' can take the values from 1 to 6, and muestra is 1 or 2. > Commonly, for each 'date' and 'id_punto' we have two samples (muestra), and > the 'flow' is also commonly the same for the two samples. > > The case is that for every 'date' I want to have the 'time series' of flows > in the way: > date 1 2 3 4 6 > 03/10/2002 606.92 323.08 440.69 393.61 5.49 > 17/10/2002 348.19 400.32 319.33 211.26 2.53 > 29/10/2002 411.57 409.66 310.37 362.20 9.36 > 14/11/2002 446.96 373.72 302.79 348.69 5.18 > 26/11/2002 381.20 386.85 307.77 240.75 6.41 > 12/12/2002 416.84 307.03 317.94 249.36 5.64 > 26/12/2002 743.04 380.30 539.94 521.60 3.69 > 09/01/2003 403.07 301.16 370.30 379.41 3.69 > ... > > where the names of the rows are the values in 'id_punto' and the values of > the data are the average (for every 'id_punto' and 'date') between the two > samples (muestra). > > In MSaccess this is solve with: > TRANSFORM avg(muestras_rambla.flow) AS avg_flow > SELECT muestras_rambla.date > FROM muestras_rambla > GROUP BY muestras_rambla.date > PIVOT muestras_rambla.id_punto; > > All I can do in postgres is : > -------- > SELECT muestras_rambla.fecha, id_punto, avg(caudal) AS avg_caudal > FROM muestras_rambla > GROUP BY fecha, id_punto > -------- > to get: > fecha | id_punto | avg_caudal > --------------------------+----------+------------------ > 03/10/2002 00:00:00 CEST | 1 | 606.924 > 03/10/2002 00:00:00 CEST | 2 | 323.079 > 03/10/2002 00:00:00 CEST | 3 | 440.690 > ... > But, how could manage this to get the 'time series' table? > > Thanks for your help and best wishes > > Javier > > ---------------------------(end of broadcast)--------------------------- > TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faqs/FAQ.html > How about: SELECT date, (SELECT s1.flow FROM muestras_rambla s1 WHERE s1.date=t1.date AND s1.id_punto=1) AS flow_1, (SELECT s2.flow FROM muestras_rambla s2 WHERE s2.date=t1.date AND s1.id_punto=2) AS flow_2, (SELECT s3.flow FROM muestras_rambla s3 WHERE s3.date=t1.date AND s1.id_punto=3) AS flow_3, (SELECT s4.flow FROM muestras_rambla s4 WHERE s4.date=t1.date AND s1.id_punto=4) AS flow_4, (SELECT s6.flow FROM muestras_rambla s6 WHERE s6.date=t1.date AND s1.id_punto=6) AS flow_6 FROM muestras_rambla t1 GROUP BY t1.date; This uses sub-queries in the Select clause to get the data in columns. There may be another way. Nick Barr
Nick Barr wrote:
>>> called "cross references querys" but I can't find my way in Postgres.
>>>
>>> My table 'muestras_rambla' is like:
>>> date | id_punto | muestra | flow | n_nitrato ...
>>> --------------------------+----------+---------+---------+-----------
>>> 06/02/2003 00:00:00 CET | 1 | 1 | 699.462 | 18.44
>>> 06/02/2003 00:00:00 CET | 1 | 2 | 699.462 | 13.79
>>> 06/02/2003 00:00:00 CET | 2 | 1 | 341.05 | 11.15
>>> 06/02/2003 00:00:00 CET | 2 | 2 | 341.05 | 17.3
>>> 06/02/2003 00:00:00 CET | 3 | 1 | 514.05 | 22.17
>>> 06/02/2003 00:00:00 CET | 3 | 2 | 514.05 | 16.62
>>> 06/02/2003 00:00:00 CET | 4 | 1 | 466.884 | 18.99
>>> 06/02/2003 00:00:00 CET | 4 | 2 | 466.884 | 19.02
>>> 06/02/2003 00:00:00 CET | 6 | 1 | 30.012 | 5.91
>>> 06/02/2003 00:00:00 CET | 6 | 2 | 30.012 | 4.15
>>> 19/02/2003 00:00:00 CET | 1 | 1 | 911.43 | 17.66
>>> 19/02/2003 00:00:00 CET | 1 | 2 | 911.43 | 9.23
>>> ...
>>>
>>> where I 'id_punto' can take the values from 1 to 6, and muestra is 1
>>> or 2. Commonly, for each 'date' and 'id_punto' we have two samples
>>> (muestra), and the 'flow' is also commonly the same for the two samples.
>>>
>>> The case is that for every 'date' I want to have the 'time series' of
>>> flows in the way:
>>> date 1 2 3 4 6
>>> 03/10/2002 606.92 323.08 440.69 393.61 5.49
>>> 17/10/2002 348.19 400.32 319.33 211.26 2.53
See the crosstab function in contrib/tablefunc. Here's what is looks
like given the above data:
select * from crosstab(
'select thedate, id_punto, avg(flow) from muestras_rambla
group by thedate, id_punto order by 1,2',
'select distinct id_punto from muestras_rambla
order by 1'
) as (thedate timestamp with time zone,
c1 float8, c2 float8, c3 float8, c4 float8, c6 float8);
thedate | c1 | c2 | c3 | c4 | c6
------------------------+---------+--------+--------+---------+--------
2003-02-05 15:00:00-08 | 699.462 | 341.05 | 514.05 | 466.884 | 30.012
2003-02-18 15:00:00-08 | 911.43 | | | |
(2 rows)
This form of crosstab() requires 7.4RC1 or a 7.3 backpatched copy from here:
http://www.joeconway.com/
HTH,
Joe