api: extend connect with fetch_schema param

[GRISHNOV]

Added support of the fetch_schema parameter, which allows to ignore schema changes on the server.

By default, it is used fetch_schema = True:

>>>importtarantool>>>conn=tarantool.Connection(host='localhost', port=3303)
>>>conn.schema_version80>>>conn.schema.schema
{257: <tarantool.schema.SchemaSpaceobjectat0x10fd21a80>, '_vinyl_deferred_delete': <tarantool.schema.SchemaSpaceobjectat0x10fd21a80>, 272: <tarantool.schema.SchemaSpaceobjectat0x10fd229b0>, ... ... , 'tester': <tarantool.schema.SchemaSpaceobjectat0x10fd22830>}
>>>tester=conn.schema.schema['tester']
>>>tester<tarantool.schema.SchemaSpaceobjectat0x10fd22830>>>>tester.
tester.aritytester.flush() tester.formattester.indexestester.nametester.schematester.sid>>>tester.format
{'id': {'name': 'id', 'type': 'unsigned', 'id': 0}, 0: {'name': 'id', 'type': 'unsigned', 'id': 0}, 'name': {'name': 'name', 'type': 'string', 'id': 1}, 1: {'name': 'name', 'type': 'string', 'id': 1}}

If the fetch_schema is specified as False, fields schema_version and schema will no longer be present in the connection object:

>>>importtarantool>>>conn=tarantool.Connection(host='localhost', port=3303, fetch_schema=False)
>>>hasattr(conn, 'schema_version')
False>>>hasattr(conn, 'schema')
False

In this case, requests to the server via the connection.update_schema will no longer be made when there is SchemaReloadException.

Closes#219

[DifferentialOrange]

In this case, requests to the server via the connection.update_schema will no longer be made when there is SchemaReloadException.

So there won't be any way to use a connection for any space operations? Only calls and evals? To be honest, I don't quite get this one without tests.

[GRISHNOV]

In this case, requests to the server via the connection.update_schema will no longer be made when there is SchemaReloadException.

So there won't be any way to use a connection for any space operations? Only calls and evals? To be honest, I don't quite get this one without tests.

I suppose, it is.
By analogy with an existing solution:

c = netbox.connect(uri, {fetch_schema = false})
c.space -- always will be nil

[GRISHNOV]

Now the connection with fetch_schema=False will throw an exception NotSupportedError when trying to call the methods replace, insert, delete, upsert, update, select.

[DifferentialOrange]

Thank you for your PR and sorry I'm a bit late (I wasn't sure that it was ready for review).

[DifferentialOrange]

You'll also need to rebase, sorry for the inconveniences

[GRISHNOV]

While working on the task, I noticed that there is an issue in the CRUD tests.
When running make test locally, in the absence of an installed crud rocks, CRUD tests freezes forever.
So, I made some changes to the CRUD instance configuration file

[DifferentialOrange]

Thank you for the updates! There are still some things we need to resolve

[DifferentialOrange]

Thank you for the updates! Let's fix remaining minor things and merge this one

[GRISHNOV]

Thanks for the feedback!

I've made all the corrections. At the moment there is a problem with tests that fall with variable probability. At the moment, it is difficult to determine the cause, but the problem, as far as I can see, is related to the update_schema call for connection pool:

# Turning the same connection into schemafull.ifmodeisnotNone:
foraddrincon.pool.keys():
con.pool[addr].conn.update_schema(con.pool[addr].conn.schema_version) # <------else:
con.update_schema(con.schema_version)

It looks like there is a desynchronization with the transmitted con.pool[addr].conn.schema_version value and the value on the server

[DifferentialOrange]

Everything seems good, I'll wait till you resolve this flaky test issue.

[DifferentialOrange]

Ping me when everything is ready