Hacker News new | ask | show | jobs
by koolba 3725 days ago
More of an ER modeler than a database designer. The interface is pretty slick/simple. Here's a sample I just cooked up: http://dbpatterns.com/documents/570699101514b428de8893a1

I can't imagine using it for anything real but it was fun to play with.

If the creator is reading this, here's a bug report: If the user hasn't hit "Save" when they export the schema, they get an empty text file and they'll think the app is broken. It should either auto-save or prompt the user "You have unsaved content, do you want to save before exporting?"

UPDATE: One more bug report, the export doesn't escape the columns names at all. It just naively adds double quotes around them. This breaks with double quotes in field names (yeah yeah which is stupid but still..):

    CREATE TABLE "foo" (
	"id" int,
	"baz" varchar(255),
	"bar" varchar(255),
	""test"" varchar(255)
    );
4 comments

The social aspects of this are awesome.

But the actual modelling functionality itself is really lacking. The lines joining the entities don't even indicate cardinality (does a ninja have many ninja_nelts? Or is it the other way around?). That's pretty much 101 for any visual database modelling.

Also the lines overlap, meaning you can't see whether ninja_weapon is related to ninja, or to ninja_belt.

I can answer these questions myself by perusing the foreign key columns - but that makes the lines redundant. Since several decades ago, Chen, Bachman, Grady Booch and a bunch of others have given ways to denote cardinality etc., and these are widely used. This tool should use one.

(apologies in advance if I have this wrong and the tool does support this but the diagram authors elected not to use them).

I found the interface very off-putting. If this is a site about exploring data models, a "public feed" and "search" is not a great way to interact with those. In fact, a timeline chronology of when these things were made seems a great way to expose mostly irrelevant information to a large number of users.

How do I browse these models? Is there a taxonomy of organization? How many models are in the system? Are there models relevant to general domains of interest to me? None of this is made apparent.

I think they're aiming at the "jsfiddle" field. I.e. Sharing ERDs and such over SO and other social sites. As opposed to them trying to be some sort of comprehensive database of standardized DB patterns.
Thanks for your sample link. It's way more interesting than the feed of model snippets in the original link.
Isn't that how you design databases? Start with an ER model? Its easy to translate to database tables.