348

What is best way to check if value is null or empty string in Postgres sql statements?

Value can be long expression so it is preferable that it is written only once in check.

Currently I'm using:

coalesce( trim(stringexpression),'')='' 

But it looks a bit ugly.

stringexpression may be char(n) column or expression containing char(n) columns with trailing spaces.

What is best way?

3
  • 3
    Using char is almost always the wrong choice due to padding (and the resulting space waste). But apart from that: I don't think there is a better solution. Commented May 20, 2014 at 17:30
  • Why ugly? Logical and legible. Commented May 20, 2014 at 17:35
  • 2
    @a_horse_with_no_name: I think there is. Commented May 20, 2014 at 18:47

11 Answers 11

532

The expression stringexpression = '' yields:

true   .. for '' (or for any string consisting of only spaces with the data type char(n))
null   .. for null
false .. for anything else

"stringexpression is either null or empty"

To check for this, use:

(stringexpression = '') IS NOT FALSE 

Or the reverse approach (may be easier to read):

(stringexpression <> '') IS NOT TRUE 

Works for any character type including char(n).
The manual about comparison operators.

Or use your original expression without trim(), which would be costly noise for char(n) (see below), or incorrect for other character types: strings consisting of only spaces would pass as empty string.

coalesce(stringexpression, '') = '' 

But the expressions at the top are faster.

"stringexpression is neither null nor empty"

Asserting the opposite is simpler:

stringexpression <> '' 

Either way, document your exact intention in an added comment if there is room for ambiguity.

About char(n)

The data type char(n) is short for character(n).
char / character are short for char(1) / character(1).
bpchar is an internal alias of character. (Think "blank-padded character".)
This data type is supported for historical reasons and for compatibility with the SQL standard, but its use is discouraged in Postgres:

In most situations text or character varying should be used instead.

Do not confuse char(n) with other, useful, character types varchar(n), varchar, text or "char" (with double-quotes).

In char(n) an empty string is not different from any other string consisting of only spaces. All of these are folded to n spaces in char(n) per definition of the type. It follows logically that the above expressions work for char(n) as well - just as much as these (which wouldn't work for other character types):

coalesce(stringexpression, ' ') = ' ' coalesce(stringexpression, '') = ' ' 

Demo

Empty string equals any string of spaces when cast to char(n):

SELECT ''::char(5) = ''::char(5) AS eq1 , ''::char(5) = ' '::char(5) AS eq2 , ''::char(5) = ' '::char(5) AS eq3; 

Result:

 eq1 | eq2 | eq3 ----+-----+---- t | t | t 

Test for "null or empty string" with char(n):

SELECT stringexpression , stringexpression = '' AS base_test , (stringexpression = '') IS NOT FALSE AS test1 , (stringexpression <> '') IS NOT TRUE AS test2 , coalesce(stringexpression, '') = '' AS coalesce1 , coalesce(stringexpression, ' ') = ' ' AS coalesce2 , coalesce(stringexpression, '') = ' ' AS coalesce3 FROM ( VALUES ('foo'::char(5)) , ('') , (' ') -- not different from '' in char(n) , (null) ) sub(stringexpression); 

Result:

 stringexpression | base_test | test1 | test2 | coalesce1 | coalesce2 | coalesce3 ------------------+-----------+-------+-------+-----------+-----------+----------- foo | f | f | f | f | f | f | t | t | t | t | t | t | t | t | t | t | t | t null | null | t | t | t | t | t 

Test for "null or empty string" with text:

SELECT stringexpression , stringexpression = '' AS base_test , (stringexpression = '') IS NOT FALSE AS test1 , (stringexpression <> '') IS NOT TRUE AS test2 , coalesce(stringexpression, '') = '' AS coalesce1 , coalesce(stringexpression, ' ') = ' ' AS coalesce2 , coalesce(stringexpression, '') = ' ' AS coalesce3 FROM ( VALUES ('foo'::text) , ('') , (' ') -- different from '' in sane character types , (null) ) sub(stringexpression); 

Result:

 stringexpression | base_test | test1 | test2 | coalesce1 | coalesce2 | coalesce3 ------------------+-----------+-------+-------+-----------+-----------+----------- foo | f | f | f | f | f | f | t | t | t | t | f | f | f | f | f | f | f | f null | null | t | t | t | t | f 

db<>fiddle here
Old sqlfiddle

Related:

Sign up to request clarification or add additional context in comments.

16 Comments

@a_horse_with_no_name: OP asks for the best way to check if value is null or empty string. The trim() call is (comparatively) expensive - and just not necessary. I added more about char(n) and "empty string".
You wrote that any string expression containing only spaces is equal to '' . Can I remove trim and use coalesce(stringexpression,'')='' to check. This looks more readable to me compared to your answer.
@Andrus: Yes, you can. I added that and some more to the answer.
select coalesce(' ', '') = '' returns false. So TRIM() is required
Using (stringexpression = '') IS NOT FALSE is a bad advice because of poor readablity. It may be confusing for the reader as to what happens when stringexpression is NULL. Using coalesce fot this is a much better idea. Readability is important. If someone misunderstands your code it may result in a bug. A lot of code is read by multiple people, with varying experience, over multiple years.
|
123

To check for null and empty:

coalesce(string, '') = '' 

To check for null, empty and spaces (trim the string)

coalesce(TRIM(string), '') = '' 

2 Comments

btw, coalesce(a, ...b) returns the first non-null value (postgresql.org/docs/8.1/functions-conditional.html)
This presumably performs a linear search and ignores the index, so you should filter based on indexed fields first.
39

Checking for the length of the string also works and is compact:

where length(stringexpression) > 0; 

6 Comments

Did you check this for the NULL case?
Yes, I did. It does not return empty nor null string fields.
If you just need to check only empty values, then try this -> where length(stringexpression) = 0;. This works for me.
I like this solution, just note that is some cases calculating exact length may be relatively expensive operation.
In postgres, length(NULL) > 0 return NULL, not False.
|
18

A lot of the answers are the shortest way, not the necessarily the best way if the column has lots of nulls. Breaking the checks up allows the optimizer to evaluate the check faster as it doesn't have to do work on the other condition.

(stringexpression IS NOT NULL AND trim(stringexpression) != '') 

The string comparison doesn't need to be evaluated since the first condition is false.

Comments

13

another way is

nullif(trim(stringExpression),'') is not null 

1 Comment

this is the most concise, works in pgsql unlike the "length function" answer
2

If there may be empty trailing spaces, probably there isn't better solution. COALESCE is just for problems like yours.

Comments

2

Something that I saw people using is stringexpression > ''. This may be not the fastest one, but happens to be one of the shortest.

Tried it on MS SQL as well as on PostgreSQL.

Comments

2

I like answer by yglodt, but calculating exact length may be expensive for big sets and big strings, so I go with:

coalesce(trim('a') > '','f') 

Comments

1

found this post looking for a solution to 'don't show me data that is '' (blank or single space char) or null'. in my case, we only want to show the user records with these values populated. i hope this response helps another looking for the same. the answers above didn't work in my case.

our app is running rails with postgres. looking at how rails builds the query for .where.not(company_website: [nil, '']) in our app, which works just fine, i can see the resulting sql statement in console.

WHERE NOT ((contacts.company_website = '' OR contacts.company_website IS NULL))

i added this bit and it works as intended.

Comments

0

My preffered way to compare nullable fields is: NULLIF(nullablefield, :ParameterValue) IS NULL AND NULLIF(:ParameterValue, nullablefield) IS NULL . This is cumbersome but is of universal use while Coalesce is impossible in some cases.

The second and inverse use of NULLIF is because "NULLIF(nullablefield, :ParameterValue) IS NULL" will always return "true" if the first parameter is null.

Comments

0

If database having large number of records then null check can take more time you can use null check in different ways like : 1) where columnname is null 2) where not exists() 3) WHERE (case when columnname is null then true end)

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.