Considering the codebase is currently hosted in a personal public repository, it may be difficult for potential contributors to find the codebase. Thus, mentioning the GitHub repository on the webpage will be helpful.

In addition to the nodes mentioned in #3 there is:

• Cross Product
• Set
• Reference
• Write
• DDL

SQL queries will probably be small enough that we don't really need these elements and they can be a bit confusing when you're not expecting them.

Currently they rely on a secret to download the backend repo. I think this is because the backend repo was originally not public. Now that the backend repo is public that secret should no longer be needed.

We could perhaps even remove the JSON text entry since JSON entry is more likely to come in via upload then it is by entering in text.

If the user enters in an SQL plan and visualizes it, then they should be able to download the plan as JSON or binary

The deploy job should skip itself if the secrets are blank. This would allow CI from remote forks to still build the site and upload it as an artifact but the jobs shouldn't attempt to deploy the site and fail.

Once a user generates a plot using a plan, they should be able to share the link to the fiddle page. The feature should store the necessary data for the plot generation, so others can access it.

Currently, if I put in the query SELECT add(l_linenumber, 2) FROM lineitem then the expression is serialized (if you click on the project node you will see this in the list of properties) as:

expressions[0] | (l_orderkey,2)

The column is wrong (l_orderkey instead of l_linenumber) and the function is missing. Ideally, it would be nicer to have:

expressions[0] | add(l_linenumber,2)

There is a placeholder ("Add a validation override level"). However, since the default override includes 1002, the user probably will not see this placeholder (unless they remove the 1002).

Once we are confident with the required security policies for a custom schema, the Show/Modify Schema option should allow modifying the existing schema.

Currently, the fiddle uses DuckDb for translating the SQL query to a Substrait plan. Will it be better to provide options for different services like Isthmus for the operation to the user?

I don't expect many people to be using the site on mobile so this is pretty low priority. Currently the fields get jumbled together and there is overlap. To start with, we should collapse the two-column display into a single column display.

Let's start with all the logical nodes. So we will need to add:

• Aggregate
• Fetch
• Sort