Responding to @sandyryza's comments

Responding to @rexledesma's comments

Forgot to run make snapshot

restructuring docs as per @rexledesma's advice

Adding to all

Refactoring the dagster GraphQL client to its own docs page

Fixing API docs issues

Addressing upstream docs comments - also (hopefully) improved error message structure for "InvalidOutputError"

Responding to more @sandyryza's comments and some upstream docs changes

rebasing

Responding to some of @sandyryza's comments

Addressing nits

nit fixes

Fixing typo

Adding in docstrings to public-facing namedtuples

Adding experimental warning to DagsterGraphQLClient

Moving DagsterGraphQLClient to the top level

Addressing comments

Modifying commit message

Modifying commit message

Modifying commit message

Adding README.md for the dagster-graphql-client CLI

rebasing

Trying installation of dagster-graphql to fix build

Rebasing

Adding dagster-graphql to tox.ini and setup.py in automation

Moving submit_pipeline_execution tests to the GraphQL python client's test suite

Fixing are_queries_compatible

Added regex check to remove redundant spacing from query strings as well

Adding checks for redundant dagster-graphql-client query snapshot calls and refactoring

Rebasing

Responding to comments - refactoring cli to dagster-graphql-client cli

Adding to bk python scripts

Adding marker to pytest.ini

Addressing comments, including seperating tests into their own tox suite

LGTM!

Also updating MySQL docs

Updating API docs to include Consolidated Sqlite Storage

Ran a make build and seemed to work for MySQL stuff

Fixing more typos

Rebase conflicts, so making sure no issues w. build

UNIQUE_TEXT => UniqueText

Adding MySQLCompatabilityTypes.UNIQUE_TEXT as a replacement for passing around UNIQUE_VARCHAR_LEN

Yup, added back pin locally and still seems to work.

In D6943#184182, @dgibson wrote:

(you could double check by leaving in the pin locally and verifying that tests still pass after changing it to scalar)

So I just checked https://www.postgresql.org/docs/9.1/sql-notify.html and verified that for a NOTIFY (registered on store_event) to be registered, LISTEN must first be called; LISTEN is first called in the pynotify loop, so all events after that will be registered & checked in the loop body. This solution looks great, since it ensures that the watcher thread is ready to listen before the event storing/notifier thread can continue.

Updating scalar call

This looks great to me, since only the PG tests have shown to be flaky right? Though honestly I'm pretty puzzled why test_event_log_storage_watch was flaky and why test_event_log_storage_two_watchers was not flaky? They're both such similar tests - the latter is just a more complex one than the former.

again

rebase

Adding @experimental tag to dagster-pandas as per @rexledesma's comment

Getting rid of D6892 & just removing signal handling logic here

rebasing

Rebasing