2 ms·

We gave the geospatial features a try with a real estate dataset that contains ~50M rows. It's crappy to say the least. Half-baked features, poor performance, r

by kijin 5y ago

We gave the geospatial features a try with a real estate dataset that contains ~50M rows. It's crappy to say the least. Half-baked features, poor performance, random crashes, barely documented gotchas with no good reason. At least the MariaDB folks try to fix the problems that are passed down to them. MySQL just seems content to tick the box, "Yeah we have that feature too, please don't switch to PostgreSQL!" and leave it at that. MySQL is okay for typical CRUD workloads with old-fashioned scalar types, but don't expect much with respect to the newer, fancier features.

datalopers 5y ago
MySQL’s spatial indexes work just fine and yield extremely low-latency queries when used correctly.
- kijin 5y ago
  Well, the last time we tried to insert a bunch of complex polygons (hundreds of points, multiple parts, etc.) into an indexed column using several concurrent transactions, MySQL crashed.
evanelias 5y ago
The list of companies using MySQL as their primary database is absolutely massive. It's a sizeable chunk of the world's economy and it's not just used for CRUD workloads. InnoDB is admittedly much better at OLTP than OLAP, but OLTP encompasses a lot more than simple CRUD workloads. > At least the MariaDB folks try to fix the problems that are passed down to them This assertion simply does not make sense. MariaDB hard-forked its codebase from MySQL's a full decade ago. They were able to use/port some open source MySQL code through ~5.7 but that seems to have stopped in recent years with MySQL 8, probably due to its major architectural changes. These days MySQL and MariaDB each have some nice features that the other one lacks. With respect to what is "passed down", by whom and to whom: the creator of MySQL is literally the chief architect at MariaDB, and many other early MySQL AB engineers work at MariaDB. Meanwhile most of MySQL's poor reputation today is a lingering effect of its early versions, largely before the Oracle acquisition of Sun. In this context, I am continually confused as to why some Postgres users have a higher opinion of MariaDB than MySQL. Edit to add: I failed to mention that I do absolutely concede that for a new application with heavy geospatial component, Postgres+PostGIS is absolutely the correct choice. InnoDB's geospatial support is comparatively much newer and less widely used.

by kijin 5y ago

datalopers 5y ago

MySQL’s spatial indexes work just fine and yield extremely low-latency queries when used correctly.

kijin 5y ago
Well, the last time we tried to insert a bunch of complex polygons (hundreds of points, multiple parts, etc.) into an indexed column using several concurrent transactions, MySQL crashed.

evanelias 5y ago

The list of companies using MySQL as their primary database is absolutely massive. It's a sizeable chunk of the world's economy and it's not just used for CRUD workloads. InnoDB is admittedly much better at OLTP than OLAP, but OLTP encompasses a lot more than simple CRUD workloads. > At least the MariaDB folks try to fix the problems that are passed down to them This assertion simply does not make sense. MariaDB hard-forked its codebase from MySQL's a full decade ago. They were able to use/port some open source MySQL code through ~5.7 but that seems to have stopped in recent years with MySQL 8, probably due to its major architectural changes. These days MySQL and MariaDB each have some nice features that the other one lacks. With respect to what is "passed down", by whom and to whom: the creator of MySQL is literally the chief architect at MariaDB, and many other early MySQL AB engineers work at MariaDB. Meanwhile most of MySQL's poor reputation today is a lingering effect of its early versions, largely before the Oracle acquisition of Sun. In this context, I am continually confused as to why some Postgres users have a higher opinion of MariaDB than MySQL. Edit to add: I failed to mention that I do absolutely concede that for a new application with heavy geospatial component, Postgres+PostGIS is absolutely the correct choice. InnoDB's geospatial support is comparatively much newer and less widely used.