title |
---|
Isnull |
isnull
checks if a value is a null
, a special kind of placeholder that's used by a database when something is missing or unknown.
isnull(text column)
You can use isnull
in custom filters, or as the condition for conditional aggregations CountIf
and SumIf
. To create a custom column using isnull
, you must combine isnull
with another function that accepts boolean values, like case
.
In Metabase tables, null
s are displayed as blank cells. Additionally, for string columns, empty strings and strings containing only whitespace characters will be displayed as blank as well.
The table below shows you examples of the output of isnull
.
Metabase shows | Database value | isnull(value) |
---|---|---|
null |
true |
|
"" (empty string) |
false * |
|
" " (whitespace) |
false |
|
kitten | "kitten" |
false |
*In Oracle and Vertica databases, empty strings are treated as nulls instead.
To create a custom column using isnull
, you must combine isnull
with another function.
For example, if you want to create a custom column that contains true
when the Discount
column is null, and false
otherwise, you can use the case expression
:
case(isnull([Discount]), true, false)
Combine isnull
with the case
expression to replace missing information with something more descriptive:
For example, you can create a new custom column that will contain "Unknown feedback"
when the original [Feedback]
column is null, and the actual feedback value when [Feedback]
is has a value. The custom expression to do it is:
case(isnull([Feedback]), "Unknown feedback.", [Feedback])
Feedback | case(isnull([Feedback]), "Unknown feedback.", [Feedback]) |
---|---|
null |
"Unknown feedback." |
"" |
"" |
"I like your style." |
"I like your style." |
Data type | Works with isnull |
---|---|
String | ✅ |
Number | ✅ |
Timestamp | ✅ |
Boolean | ✅ |
JSON | ✅ |
- In Metabase, you must combine
isnull
with another expression that accepts boolean arguments (i.e.,true
orfalse
). isnull
only accepts one value at a time. If you need to deal with blank cells across multiple columns, see the coalesce expression.- If
isnull
doesn't seem to do anything to your blank cells, you might have empty strings. Try theisempty
expression instead.
This section covers functions and formulas that can be used interchangeably with the Metabase isnull
expression, with notes on how to choose the best option for your use case.
All examples below use the table from the Replacing null values example:
Feedback | case(isnull([Feedback]), "Unknown feedback.", [Feedback]) |
---|---|
null |
"Unknown feedback." |
"" |
"" |
"I like your style." |
"I like your style." |
In most cases (unless you're using a NoSQL database), questions created from the query builder are converted into SQL queries that run against your database or data warehouse.
CASE WHEN Feedback IS NULL THEN "Unknown feedback",
ELSE Feedback END
is equivalent to the Metabase isnull
expression:
case(isnull([Feedback]), "Unknown feedback.", [Feedback])
Spreadsheet #N/A
s are the equivalent of database null
s (placeholders for "unknown" or "missing" information).
Assuming our sample feedback column is in a spreadsheet where "Feedback" is in column A, then the formula
=IF(ISNA(A2), "Unknown feedback.", A2)
is equivalent to the Metabase isnull
expression:
case(isnull([Feedback]), "Unknown feedback.", [Feedback])
Numpy and pandas use NaN
s or NA
s instead of null
s.
Assuming our sample feedback column is in a dataframe column called df["Feedback"]
:
df["Custom Column"] = np.where(df["Feedback"].isnull(), "Unknown feedback.", df["Feedback"])
is equivalent to the Metabase isnull
expression:
case(isnull([Feedback]), "Unknown feedback.", [Feedback])