如何使用SQL LIKE运算符在列中搜索精确模式？

Question

I have a table which is like below

    date    |                                                   tags                                                   
------------+----------------------------------------------------------------------------------------------------------
 2018-10-24 | {"table": "bank_trans", "metric": "withdrawal", "location": "UK"}
 2018-10-24 | {"table": "bank_trans", "metric": "balance", "account_id": "477", "location": "ny", "country": "USA"}
 2018-10-24 | {"table": "bank_trans", "metric": "deposit", "location": "blr", "country": "IND"}
 2018-11-02 | {"table": "bank_trans", "metric": "balance", "account_id": "477"}

If I want a particular row that contains a search pattern something like below

select date, tags
from webhook_forecastmodel
where tags LIKE '%"table": "bank_trans"%' AND
      tags LIKE '%"metric": "balance"%' AND
      tags LIKE '%"account_id": "477"%';

In this case, I get back two results

    date    |                                                   tags                                                   
------------+----------------------------------------------------------------------------------------------------------

 2018-10-24 | {"table": "bank_trans", "metric": "balance", "account_id": "477", "location": "ny", "country": "USA"}

 2018-11-02 | {"table": "bank_trans", "metric": "balance", "account_id": "477"}

I understand the SQL query returns me the rows where the pattern matches.

But I only want the row that is exactly mentioned in the LIKE search pattern which is "table": "bank_trans" , "metric": "balance" and "account_id": "477" which leaves us with only one row

 2018-11-02 | {"table": "bank_trans", "metric": "balance", "account_id": "477"}

Is there any possible way that this could be achieved?

Answer 1

You have a really bad data structure. Postgres supports a JSON type, so you could use that. However, even with a JSON type, your question would be a bit challenging.

More importantly, you seem to want this information in columns so you should put it there.

Given the constraints in your question, there is a simple solution if I assume that , does not appear inside the key/value pairs. If so:

select date, tags
from webhook_forecastmodel
where tags LIKE '%"table": "bank_trans"%' AND
      tags LIKE '%"metric": "balance"%' AND
      tags LIKE '%"account_id": "477"%' AND
      tags NOT LIKE '%,%,%,%';  -- not more than three key/value pairs

Answer 2

UPDATE : this question assumes an up-to-date Postgres version. It will not work on the outdated and no longer maintained version 9.2, but I'm leaving it here anyway for reference.

---

As mentioned in the comments don't use LIKE, use JSON functions . In order to be able to do that, you have to cast the value:

select date, tags
from webhook_forecastmodel
where tags::jsonb @> '{"table": "bank_trans"}'::jsonb 
  AND tags::jsonb @> '{"metric": "balance"}'::jsonb 
  AND tags::jsonb @> '{"account_id": "477"'}::jsonb;

The @> operator checks if the value on the left side contains the key/value pair on the right side.

---

The above would also return rows that contain more than those key/value pairs. If you want those that contain exactly those key/value pairs, use =

select date, tags
from webhook_forecastmodel
where tags::jsonb = '{"table": "bank_trans", 
                      "metric": "balance", 
                      "account_id": "477"}'::jsonb;

The jsonb data type normalizes the key/value pairs, so the order of the keys is irrelevant and the = comparison will work correctly.

Online example: https://rextester.com/LYXHUC20162

In the online example I have a different order of keys in the tags column compared to those used for the = operator to demonstrate that JSONB normalizes the JSON representation.

---

Given the nature of your data, it's probably better to define the column as jsonb to avoid all the casting.