使用 Javascript 从 BigQuery 中的上一行和同一列计算

Question

i need to calculate pending_principal from the previous raw, is there any I can do this directly in SQL? I'm using BigQuery and Javascript.

sample Data

Date	Rawnumb	late_fee	interest	Pending_principal
2020-01-01	1	0	100000	1000000
2020-01-02	2	null	150000	null
2020-01-03	3	null	200000	null
2020-01-04	4	null	250000	null
2020-01-05	1	100000	300000	1000000
2020-01-06	2	null	900000	null

i want to calculate pending_principal and late_fee which contains null values

The logic for late_fee if rownumb=1 late fee already exist on table but if rownum is not 1 the logic is:

late_fee=5% * previous row pending_principal

The logic for pending_principal if rownumb=1 pending_principal already exist on table but if rownum is not 1 the logic is:

pending_principal=previous pending_principal+late_fee+interest

for example on 2020-01-02

late_fee=5%*1.000.000=50.000

pending_principal=1.000.000+50.000+150.000=1.200.000

i write the query:

CREATE TEMP FUNCTION udf_calc(x ARRAY<STRUCT<rownum INT64, late_fee INT64, interest INT64, pending_principal INT64>>)
RETURNS STRUCT<rownum INT64,late_fee INT64, interest INT64, pending_principal INT64>
LANGUAGE js
AS """
  var vrownum = 0;
  var vlate_fee = 0;
  var vinterest = 0;
  var vpending_principal = 0;
  for (var row of x)
  {
    if (vrownum == 1) {
      vlate_fee=row.late_fee
    }
    else {vlate_fee = parseInt(vpending_principal) * 0.05}
    ;
    if (vrownum === 1) {
      vpending_principal = row.pending_principal;
    }
    else {
      vpending_principal = parseInt(vpending_principal) + parseInt(vlate_fee) + parseInt(row.interest);
    }
    vinterest = row.interest;
    vrownum = row.rownum;
  }
  r = {rownum:vrownum,late_fee:vlate_fee, interest:vinterest, pending_principal:vpending_principal};
  return r;
""";

WITH mytable AS (
  SELECT date '2020-01-01' as date, 1 as rownum , 0 as late_fee, 100000 as interest, 1000000 as pending_principal UNION ALL
  SELECT date '2020-01-02',2 , null, 150000, null UNION ALL
  SELECT date '2020-01-03',3 , null, 200000, null UNION ALL
  SELECT date '2020-01-04',4 ,null, 250000, null UNION ALL
  SELECT date '2020-01-05',1 , 100000 , 300000, 100000 UNION ALL
  SELECT date '2020-01-06',2 ,  null, 900000, null
)
select date,
  udf_calc(array_agg(STRUCT(rownum, late_fee, interest, pending_principal)) over (order by date rows unbounded preceding)).*
from mytable

but the result is not correct

Date	rownum	late_fee	interest	Pending_principal
2020-01-01	1	0	0	1000000
2020-01-02	2	50000	150000	1200000
2020-01-03	3	60000	200000	1460000
2020-01-04	4	73000	250000	1783000
2020-01-05	1	89150	300000	2172150
2020-01-06	2	108608	900000	3180757

i expect the result is

Date	rownum	late_fee	interest	Pending_principal
2020-01-01	1	0	0	1000000
2020-01-02	2	50000	150000	1200000
2020-01-03	3	60000	200000	1460000
2020-01-04	4	73000	250000	1783000
2020-01-05	1	100000	300000	1000000
2020-01-06	2	50000	900000	1950000

i think my script didnt read the condition if rownum==1

Is that possible in some way?

Answer 1

It is easier to split rows into groups with something like group_num :

CREATE TEMP FUNCTION udf_calc(x ARRAY<STRUCT<late_fee INT64, interest INT64, pending_principal INT64>>)
RETURNS STRUCT<late_fee INT64, interest INT64, pending_principal INT64>
LANGUAGE js
AS """
  var vlate_fee = 0;
  var vinterest = 0;
  var vpending_principal = 0;
  for (var row of x)
  {
    vlate_fee = parseInt(vpending_principal) * 0.05;
    if (vpending_principal === 0) {
      vpending_principal = row.pending_principal;
    }
    else {
      vpending_principal = parseInt(vpending_principal) + parseInt(vlate_fee) + parseInt(row.interest);
    }
    vinterest = row.interest;
  }
  r = {late_fee:vlate_fee, interest:vinterest, pending_principal:vpending_principal};
  return r;
""";

WITH mytable AS (
  SELECT date '2020-01-01' as date, 1 as group_num , 0 as late_fee, 100000 as interest, 1000000 as pending_principal UNION ALL
  SELECT date '2020-01-02',1 , null, 150000, null UNION ALL
  SELECT date '2020-01-03',1 , null, 200000, null UNION ALL
  SELECT date '2020-01-04',1 ,null, 250000, null UNION ALL
  SELECT date '2020-01-05',2 , 100000 , 300000, 1000000 UNION ALL
  SELECT date '2020-01-06',2 ,  null, 900000, null
)
select date,
  udf_calc(array_agg(STRUCT(late_fee, interest, pending_principal)) over (partition by group_num order by date rows unbounded preceding)).*
from mytable