Merge pull request #172 from awslabs/sql_sample_apps

Sql sample apps

sample_apps/sql/last_value_fill_forward/README.md

@@ -2,9 +2,11 @@
 
 ## Query uses multiple steps
 
-1. create time sequence (not limited to just 10,000 data points)
-2. select raw data binned at same intervals
-3. join time sequence with raw data as data set that contains NULL values now
+1. create time sequence `time_seq_only` (not limited to just 10,000 data points)
+2. get all distinct device ids (in this data example as gpio) as `distinct_gpio`
+2. duplicate time sequence for device id's to `time_seq_with_gpio` (in this example gpio channels) to allow for each channel to be filled individually
+2. select raw data binned at same intervals as `raw_pos`
+3. join time sequence `time_seq_with_gpio` with raw data `raw_pos` as data set that contains NULL values now
 4. use LAST_VALUE in filled dataset, this query lists justs 2 measures: orignal temperature that can contain NULL and the filled column
 
 The result set shows both original value containing NULL and the filled value in a separate column

sample_apps/sql/last_value_fill_forward/query_example_fill_forward.sql

@@ -1,25 +1,32 @@
-with time_seq as ( -- Timesequence is 30,240 data points, starting 2 weeks ago of total 3 weeks (1 week in future) with 1 min intervals
+with time_seq_only as ( -- Timesequence is 30,240 data points, starting 2 weeks ago of total 3 weeks (1 week in future) with 1 min intervals
     select
     date_add('day', day,
     date_add('hour', hour,
     date_add('second', second, bin('2023-09-20 17:12:22.958000000',1m)))) as time
     from unnest(sequence(0,3540,60)) t(second) cross join unnest (sequence(0, 23)) as t(hour) cross join unnest (sequence(0, 20)) as t(day)
 	order by day, hour, second
 ),
+distinct_gpio as ( -- each device identified by gpio needs its own time sequence, this query gets all device ids
+    select distinct(gpio) as gpio from "amazon-timestream-tools"."sensordata"
+),
+time_seq_with_gpio as ( -- multiple time sequences, one for each device identified by gpio
+    select time, gpio from time_seq_only join distinct_gpio on true
+),
 raw_pos as (
 SELECT bin(time, 1m) as p_time,
    avg(temperature) as temperature,
    gpio
 FROM "amazon-timestream-tools"."sensordata" -- adjust if data is loaded to different table
    where time between '2023-09-20 17:12:22.958000000' and now() -- sample data set contains data from 09/20/2023
-   and gpio = '22'
+   -- and gpio = '22'
 GROUP BY gpio, bin(time, 1m)
 ),
 -- dataset contains missing records as just symbol (key), timestamp and all other columns are null
 dataset as (
-select '22' as gpio, bin(time, 1m) as time, temperature from time_seq
+select time_seq_with_gpio.gpio, bin(time, 1m) as time, temperature from time_seq_with_gpio
 left join raw_pos
-on time_seq.time = raw_pos.p_time
+on time_seq_with_gpio.time = raw_pos.p_time
+    and time_seq_with_gpio.gpio = raw_pos.gpio
 ),
 filled_set as (
   SELECT
@@ -35,4 +42,4 @@ filled_set as (
 )
 select * from filled_set
 -- select * from dataset -- use this line to review original data with containing gaps
-order by time
+order by gpio, time