- Rule of thumb: Avoid filtering loops
- Use ORM
- Complex queries
Rule of thumb: Avoid filtering loops
SQL/Databases are good at 2 jobs: storing data. filtering data as fast as possible. You’re for loop cannot compete with SQL.
The performance of your program will be much faster if you let SQL manage the filtering of data. Plus it will also result in a smaller line of code.
Try to filter with smaller queries with a smaller number of columns before fetching all the columns. Smaller the data, the faster will be the data transfer. You will find a better example of this in the section of Complex Queries bellow.
For Spring (Java), ideally, use JPA method query things.***
_Tips: Use tools to make Entities POJO
For Python, SQLAlchemy is popular and preferred or Django ORM for Django. (I have less experience with this)
Native queries are faster than usual ones but use the sparingly.
In the case of Java, you can also JPA specific things like Specifications or Criteria API.
Honestly, I haven’t tried them I know they exist. I have seen them and found that the code looked really ugly
I do find Django ORM to be easier than Spring JPA (you can pass arguments rather than always create a separate function).
Break down big-bang (complex) queries into multiple simpler queries
note: This might be very controversial and I have not seen this talked about anywhere or used in other people’s code. You are not likely to use this unless you are like me (bad at ORMs).
Many times you have a complex feature like search and you want to filter by multiple parameters.
What usually ends up happening you end up with a custom query generator or Criteria API (Java) or a query which is too long. Especially if this query requires additional features like ordering and pagination.
I think a simpler solution is to break it into smaller queries. I give a simplified example below.
Note that this is a much simpler version of what you would encounter but I think it will make the point.
There are 3 tables:
Suppose you want to find who are male grandparents of children whose name start with ‘A’
You can go for some complex query like
These 33 lines can quickly increase in size when you have too many tables. Multiple shorter queries are always much better than one big bang query.
I’ve seen queries which go beyond 100 lines 😖
Making changes in such large 100+ line queries is a pain.
(Even when writing the 33 line query above, I introduced bug by not writing join on lines 30 and 31. writing smaller queries helped me debug.)
You could break into something simpler queries like:
Then code the method yourself to call these in a certain sequence:
You will find that even though your line of code has increased, the modularity of code has also increased and these functions can be arranged in any way you like.
Next: Remove unnecessary code
Index: Tips on writing Clean Code