The NoSQL trend was so terrible. Anyone starting out right in that time frame where mongo and other NoSQL DBs were getting popular was really done a disservice.
I sit in design meetings all the time where people with <5 years experience go out of their way to avoid using a relational database for relational data because "SQL is slow". They will fight tooth and nail, shoe-horning features in to the application which are trivial to do with a single SQL command.
I helped out on one project lead by a few younger devs who chose FireStore over CloudSQL for "performance reasons" (for an in-house tool). They had to do a pretty major rewrite after only a few weeks once they got around to deleting, because one of their design requirements was to be able to delete thousands of records; a trivial operation in SQL, but with FireStore, deleting records requires:
> To delete an entire collection or subcollection in Cloud Firestore, retrieve all the documents within the collection or subcollection and delete them. If you have larger collections, you may want to delete the documents in smaller batches to avoid out-of-memory errors. Repeat the process until you've deleted the entire collection or subcollection.
> Deleting a collection requires coordinating an unbounded number of individual delete requests.
Turns out, once they started needing to regularly delete thousands-millions of records, the process could run all night. Luckily, moving over to CloudSQL didn't take very long...
> I sit in design meetings all the time where people with <5 years experience go out of their way to avoid using a relational database for relational data because "SQL is slow"
I mean, this is just dumb. I have less than 5 years experience and I understand that SQL isn't "slow", there are just different tradeoffs between SQL and NoSQL databases and you have to pick the right tool for the job.
I sit in design meetings all the time where people with <5 years experience go out of their way to avoid using a relational database for relational data because "SQL is slow". They will fight tooth and nail, shoe-horning features in to the application which are trivial to do with a single SQL command.
I helped out on one project lead by a few younger devs who chose FireStore over CloudSQL for "performance reasons" (for an in-house tool). They had to do a pretty major rewrite after only a few weeks once they got around to deleting, because one of their design requirements was to be able to delete thousands of records; a trivial operation in SQL, but with FireStore, deleting records requires:
> To delete an entire collection or subcollection in Cloud Firestore, retrieve all the documents within the collection or subcollection and delete them. If you have larger collections, you may want to delete the documents in smaller batches to avoid out-of-memory errors. Repeat the process until you've deleted the entire collection or subcollection.
> Deleting a collection requires coordinating an unbounded number of individual delete requests.
Turns out, once they started needing to regularly delete thousands-millions of records, the process could run all night. Luckily, moving over to CloudSQL didn't take very long...