Why comparison and conditional aggregation fails in BigQuery

Why did comparison and conditional aggregation fail in BigQuery ?

BigQuery(BQ) is very useful for data analysis or processing.

It is good at handling huge data. It returns summary result in short time.

When we handle big data, sometimes we want to extract or summarise data that meets specific conditions.

Usually we use comparison operator like below.

WHERE COLUMN1 = "A"

SUM(IF(COLUMN1 = "A",1,0)

But it fails in specific case.

Why didn't it work.

So today I introduce about "Why comparison and conditional aggregation fails in BigQuery".

Author

Mid-carieer engineer (AI, system). Good at Python and SQL.

Data

First, prepare data.

Import this csv file as null_sample table.

col1,col2,col3
a,b,c
,b,c
,,c
a,b,
a,,

Then you can see table like below.

Comparison and conditional aggregation

They are fine example of comparison and conditional aggregation.

We use operator =.

SELECT * FROM test.null_sample
WHERE col1 = "a"

In order to get conditional sum, we can use SUM and IF function.

SELECT
SUM(IF(col1="a",1,0)) as sum_col1_a
FROM test.null_sample

Example of failure

Then they are examples of failure.

They use != and its comparison does not work.

SELECT * FROM test.null_sample
WHERE col1 != "a"

Even if conditional aggregation, it fails like below.

SELECT
SUM(IF(col1="a",1,0)) as sum_col1_a,
SUM(IF(col1!="a",1,0)) as sum_col1_not_a,
SUM(IF(col2!="a",1,0)) as sum_col2_not_a,
SUM(IF(col3!="a",1,0)) as sum_col3_not_a
FROM test.null_sample

Sample table has 5 records.

In the table, about non- a count, it should be 2 in col1.

And it should be 5 in col2 or col3.

Why did it fail to aggregate.

Why comparison and conditional aggregation fails in BigQuery

The reason why comparison and conditional aggregation fails in BigQuery is null.

null is special that does not return true to both = nor !=.

So if you want to compare data,you should replace null with other value by IFNULL or COALESCE.

SELECT * FROM test.null_sample
WHERE IFNULL(col1,"") != "a"

For conditional aggregation, we should use IFNULL or COALESCE.

SELECT
SUM(IF(IFNULL(col1,"")="a",1,0)) as sum_col1_a,
SUM(IF(IFNULL(col1,"")!="a",1,0)) as sum_col1_not_a,
SUM(IF(IFNULL(col2,"")!="a",1,0)) as sum_col2_not_a,
SUM(IF(IFNULL(col3,"")!="a",1,0)) as sum_col3_not_a
FROM test.null_sample

In this case we got correct non-a records and non-a count.

Conclusion

Today I explained about "Why comparison and conditional aggregation fails in BigQuery".

The reason why comparison and conditional aggregation fails in BigQuery is null.

It fails when we try to compare null directly.

Solution is this.

null is very complicated.

There are some other articles about BigQuey.

If you interested in them, please read them.