issue_comments

I'm going to have the .add_database() method select the name used in the path, de-duping against any existing names. It will then set database.name to that so that the database has access to its own name.

There's some messy code that needs fixing here. The datasette.databases dictionary right now has a key that corresponds to the /_internal URL in the path, and a value that's a Database() object. BUT... the Database() object doesn't know what its key is.

While fixing this I should fix the issue where Datasette gets confused by multiple databases with the same stem: https://github.com/simonw/datasette/issues/509

Fun query: sql select table_name, group_concat(name, ', ') from columns group by database_name, table_name https://latest.datasette.io/_internal?sql=select+table_name%2C+group_concat%28name%2C+%27%2C+%27%29+from+columns+group+by+database_name%2C+table_name

I removed that MEMORY object() in dcdfb2c301341d45b66683e3e3be72f9c7585b2f

datasette-graphql returns an error due to this issue:

On the console: INFO: 127.0.0.1:63116 - "POST /graphql/_internal HTTP/1.1" 500 Internal Server Error Traceback (most recent call last): File "/Users/simon/.local/share/virtualenvs/datasette-AWNrQs95/lib/python3.8/site-packages/promise/promise.py", line 844, in handle_future_result resolve(future.result()) File "/Users/simon/.local/share/virtualenvs/datasette-AWNrQs95/lib/python3.8/site-packages/datasette_graphql/utils.py", line 603, in resolve_table data, _, _ = await view.data( File "/Users/simon/Dropbox/Development/datasette/datasette/views/table.py", line 304, in data db = self.ds.databases[database] graphql.error.located_error.GraphQLLocatedError: ':memory:c6dd5abe1a757a7de00d99b699175bd33d9a575f05b5751bf856b8656fb07edd'

A Database should have a .name which is unique across the Datasette instance and is used in the URL. The path should be optional, only set for file databases. A new .memory_name property can be used for shared memory databases.

Database internal class is documented here: https://docs.datasette.io/en/latest/internals.html#database-class

Here's the commit where I added it. https://github.com/simonw/datasette/commit/9743e1d91b5f0a2b3c1c0bd6ffce8739341f43c4 - I didn't yet have the .add_database() mechanism. Today the MEMORY object bit is no longer needed.

The code for building up that .databases dictionary is a bit convoluted. Here's the code that adds a :memory: database if the user specified --memory OR if there are no files to be attached:

https://github.com/simonw/datasette/blob/ebc7aa287c99fe6114b79aeab8efb8d4489a6182/datasette/app.py#L221-L241

I'm not sure why I wrote it this way, instead of just calling .add_database(":memory:", Database(..., is_memory=True).