Re:The Marketing Behind MongoDB
by DaiPlusPlus on Tuesday August 29, 2017 @04:01PM
(#15127173)
Adding a non-NULL column to an existing table with millions of rows, and then updating the entity/model objects upwards towards the UI is non-trivial - I find myself "queuing up" planned database schema changes and applying them in a batch.
What I'd love is a traditional database system with the concept of "cheap columns" that store "just data" scalar values, which can be easily created, dropped and manipulated - they're addressable so they can be used in queries and statistics/analytics, but the values themselves won't need indexes or referential-integrity - and for an ORM and Scaffolding system to support them - that way I won't pull out my hair when the client says one day "let's go from a single column for the customer's name to having separate title/first/middle/last/suffix columns"... and back again the week after.
So something like PostgreSQL `json` and `jsonb` types? You can select them out into columns, so they should play nicely with ORMs.
https://www.postgresql.org/docs/current/static/datatype-json...
Re:The Marketing Behind MongoDB
by solidsnack9000 on Wednesday August 30, 2017 @02:02AM
(#15129998)
It seems like the "column families" of Cassandra would appeal to you.
Schemaless db is great feature. Joins can be done through an orm like Laravel's eloquent which allows you to join mysql to mongo using a cross-join table on the mysql side.
Adding columns in an mpv product is trivial. Getting a stream of unstructured data and being able to save the different data to a document that is part of the same collection is an amazing feature that is needed still.