Data warehouse feedback

[annikaschmid]

Data warehouse

UX / UI

High

• (fix(data-warehouse): data warehouse scene ui updates #21189) Show 'last updated/refreshed at' for each table

Medium

• (fix(data-warehouse): data warehouse scene ui updates #21189) Group tables by source and add a sub headline so that the list is easier to browse. E.g. “Hubspot”, “Stripe”
• (fix(data-warehouse): data warehouse scene ui updates #21189) Better loading state for external tables. Right now it takes 5+ seconds for them to load, and in the meantime looks like this:

• (fix(data-warehouse): data warehouse scene ui updates #21189 )Clicking a view at the bottom of the main data warehouse page seems like it doesn’t do anything, because the table schema is outside of the visible area on the screen. I think it would be cool if the page would automatically scroll to the top when a table is selected, or the list of tables is scrollable while the main area of the screen containing the table schema is fixed
• After creating and saving a join, I can edit it and rename the field name, but clicking the ‘save’ button doesn’t do anything

Low

• Once data warehouse it out of beta, consider removing the 'Database' tab in data management, or link to data warehouse (as it's the same
• Some tables are in the list more than once, e.g. prod_hubspot_companies. I am assuming this is from us testing connections, but in the future, we should prevent users from being able to add multiple tables with the same name (is this still an issue? This should've been fixed a few weeks ago with: fix(data-warehouse): add schema id for validation step and use snake case #20840)
• (chore: replace remaining prompts #21218) "Save as view" popup still has the old PostHog UI

• We allow 'view' names with spaces, and then throw a toast error on saving. It would be better if we had inline validation

• (chore(data-warehouse): fix long table names #21309) When the 'view' name is very long, the UI doesn’t look good:

• (chore(data-warehouse): fix long table names #21309) Same issue in the 'join' popup. I would expect if the source table name is even longer, it wouldn’t look good.

• (chore(data-warehouse): fix long table names #21309) In the list of joins, the description/name is interactive and can be clicked, but clicking it doesn’t do anything

• The list of tables when creating a join can get very long and cannot be searched. There is also no visual differentiation between views and tables. (Not sure we need the second if we have the first, it just took me quite long to find the view I wanted to select, which is much easier to spot on the data warehouse first page)

Feature requests

High

• (feat(data-warehouse): editing views #21261) Editing the query of a view after it has been created
• (chore(data-warehouse): joins on views #21151) Using aliases (e.g. JSONExtract) in Joins. Reported here.

Open discussions

• billing_customer can't be loaded with all columns because "schema is broken"
• Setting a ‘field name’ when creating a join was a bit unclear to me what this is doing, and then how to use it. Docs might help too. Reported here.
• When to create a view vs. using a HogQL expression for a join. Eric: “At some point we need a better modeling system.”

HogQL

UX/UI

High

• Better placeholder (cached data?) when a query is being recalculated and not in edit mode. Right now, the screen is just blank, when a query is loading for 15s you don't even see anything is happening in the background

Medium

• The error messages are really tricky to interact with, e.g. sometimes there is a 2mm wide red squiggly line, that is really hard to hit with the mouse. Once the popover opens, it says to hit ‘option’ + F8 to view the problem, which doesn’t seem to work for me

Feature requests

High

• Setting variables in the query, e.g. for date or product
• Turn cells into links and using the cell content as a variable to link to another table
  -> We use these two features heavily in Metabase, and this is the main thing we are missing to move go/revenue from Metabase to PostHog. e.g. linking stripe_customer_id from this graph to this graph

Medium

• Error messages don’t reference the line/position an error occurred in (after the query has completed, this works for inline errors), and syntax errors aren’t specifically called out. Some of my errors are the result of a missing ',' and in Metabase, this is much more obvious. In PostHog, it’s quite unclear whether the problem is with the column/data/syntax, when in fact it’s a syntax error

Low

• Conditional highlighting of cells, e.g. when an invoice isn’t complete yet (Metabase has this)

[annikaschmid]

@EDsCODE Do you have a GitHub issues label you use for data warehouse? Otherwise I can create a new one, e.g. feature/data-warehouse

[mariusandra]

We allow 'view' names with spaces, and then throw a toast error on saving. It would be better if we had inline validation

Why not allow spaces everywhere? It looks nicer, and feels more natural to most people. HogQL has two ways you escape spaces in identifiers (double and back quotes), so there's no technical limitation here.

[EDsCODE]

@annikaschmid is this still a problem? I can't repro it

After creating and saving a join, I can edit it and rename the field name, but clicking the ‘save’ button doesn’t do anything