Preview upcoming changes to the GUI query builder (updated)

Now I have finally had the time to try out the new UI and have some feedback!

  1. I (as more people have pointed out) miss the “Ask a question” button. I would rather hide “Write SQL” in the menu, than “New custom question”.
  2. To highlight “Browse data” in the header feels unnecessary imo. We are also having a lot of tables and that feature is not something we use that often.
  3. I love how the result is auto updated as you compose your question!
  4. I like the new design of the advanced query editor, but I think the new design of the simpler editor is less intuitive than before. I really enjoyed the “from left to right” workflow when asking my questions. Now, when I switch from the advanced editor, to the simple interface, I get the filter and summarize buttons to the right, next to the “Save” button and when I add, for example a filter, it is added to the left. If I then want to change the filter, I press the left filter button, and can change that filter. If I want to add another filter, I need to do that from the right filter button. I would much rather be able to add another filter directly in the header of my question (similar to before).
  5. I also think that the “Summarize” concept is less intuitive than before when it was split into “View” and “Group by” and the workflow in the simple interface for choosing what to view and group by is more complex than before. You can’t see what has been selected for “View” and “Summarize by” as you can when you have selected a filter without having to press the “Summarize” button again. Why isn’t that added as buttons next to the filter button to the upper left?
  6. I think the toggle for switching to the advanced editor is misplaced to the upper right. I would rather have that to the upper left.
  7. It would be great to have a setting for the default editor (advanced or simple) used when asking questions, on a user level. Some users in my organizations are really non-technical and this release have made it more difficult for them to use Metabase.

To summarize: I really like the concept of having advanced/simple mode and can really see the possibilities further on and like the UI for the advanced editor, but the changes to the UI and workflow using the simple editor has made it more complex to use Metabase.

1 Like

Awesomeness preview2 !!! Just ready for 4th July :fireworks:
Really nice to see a lot of tiny, but impactful, adjustments since preview1.
Only did a quick-spin.

  1. What was the reason behind making the Editor-button only an icon? Given that it’s really impressive what you can do there, not sure why it’s getting “tucked away”.
  2. When clicking the Filter-button, and it shows side-panel, it is a little misleading to both have “Update filter” and “Done” buttons. Clicking done, doesn’t apply the filter.
  3. If a user only has data access, but no sql, then it still shows “Write SQL”
1 Like

Just spun up 0.33 preview 2, have spent less than 10 minutes with it, but loving it already. Definitely seeing how this will make it easier for non-technical users to interact with data more thoroughly :metal:

Hi,

We've updated to the preview, and it looks good though I haven't spent much time with it yet. I've run into what looks like a bug and I don't think it's mentioned above, but apologies if i'm wrong there:

If you do a join using the new query builder, then the joined values appear in the results table. If you click on one of the values, and say you want to filter to show only rows that have that value greater than 0, you get an error like the attached screenshot. I tried it with a different pair of tables and got the same result (filtering on text property equality).

Kind Regards,
Mat Bailey

Hi. I’m a new user. Loving the GUI builder. But would there be a possibility to filter between fields? field1 <> field2, and etc. using the GUI?

@laurenceco It will likely not be a part of 0.33, but go and upvote by clicking :+1: on the first post:
https://github.com/metabase/metabase/issues/1834

Hi team Metabase.
Let me start by saying I love your work and have been using your visualising tool for about 2 years.
I have setup the beta metabase app and we are in the process trying it out.
One killer feature that we cannot seem to find is, allowing Dashboard filters based on the Segments created against a table - or are we just missing something?

Example we create multiple Segments for job_role groups against the table and then create a widget to go onto the dash board, but the filter does not allow us to use the segments explicitly as a choice. Is this something that is being actively considered or is it just too much of a job.

Thanks again for all the hard work.

@chrisb It’s likely not going to be part of 0.33, but go and upvote by clicking :+1: on the first post:
https://github.com/metabase/metabase/issues/3015

@nmg Glad to hear you’re loving it! In using the simple summarizing and filtering mode (rather than the custom query editor), are you finding it feels clear enough to tell what it is you’re looking at? (I.e. what the metric(s) and breakout(s) are.) I’d be curious to hear if it also seems clear to less technical users you might have shown it to. We’re trying to let the page title, as well as chart legends and axes, or table column headings to do the heavy lifting when it comes to communicating the metric(s) and breakout(s) of a question as a trade off to make the default view of the page simpler.

1 Like

Small findings (preview2):

  • Having multiple SQL variables breaks the UI completely (more than 5 was enough on my resolution)
  • In Editor, the Sort section doesn’t use the new names (i.e “Sum of Total”) used in column name, just “Sum”, so it’s difficult to figure out what column you’re sorting, when having multiple “Sum of…”
  • In Editor, the group-by section doesn’t show the binning until clicked.
  • Even after specifying visualization (it auto’ed to Line and I changed to Table), then when adding another metric, it auto to Line again.
1 Like

Some thoughts/comment on V2:

  1. Would it be possible to continue to use the current “simple” layout as the default, with an option to select “advance query” mode. While I get the argument for additional functionality, the simplicity of the current design is very appealing to non techie folk.

  2. There are a number of different design paradigms for building an advanced adhoc query interface that might be informative - for example https://www.researchgate.net/figure/An-i2b2-multi-group-joint-query-interface_fig4_316332892. Though this does rely on a predefined data model. Another thought would be to explore a query builder based more along the lines of a visual workflow builder such as you would find in tools like Pentaho - essentially morphing to a light weight “etl” workflow builder. Such an interface in my opinion, provides better legibility and comprehension but also I would argue provides a platform to expand the system to integrate external data sources such as a web services at some point in the future.

  3. Table joining - It seems like there is no restriction on what columns could be used to perform a join. The Likelihood of a user joining large tables on non indexed columns seems high with the resultant performance issues. Do we want the end users to become pseudo dba’s?

  4. I wonder if it might make more sense to focus the advanced query builder on allowing the user to build queries by joining “questions” rather than the underlying tables - relying on an enhanced data model to take care of the logic (including selecting indexed columns) to optimize the query and insulate the user from the “sausage making” query logic if possible (the i2b2 app mentioned in 2 does that quite well).

  5. The “Custom Question” is hidden in a drop down is there a reason for this? Seems like there is real-estate to render on the UI or at least I’d swap “Browse Data” with Custom Question.

  6. When I “lasso” a set of points in a chart, the UI attempts to zoom on those points but fails and displays a blank chart.

  7. Although “Allow using a saved question as the source for other queries?” is enabled I dont seem to be able to select a previous question as a data source.

  8. supporting “EXISTS” and “NOT EXISTS” functionality in the query builder, ideally leveraging preexisting questions in the logic - e.g. select * from TableX where not exists in QuestionY join X.id = Y.id…

  9. Supporting date operations on the join comparator - x.birth_date between y.start_date and y.end_date

  10. Replace “geek speak” such as “Left outer join” with text that a non-techie would understand

  11. I find jumping from the top right options - Filter, summarize etc. to the bottom left Visualize, Settings non-initutive. Any possibility of moving the latter up beside the Summarize option? Similarliy have the visualization and setting windows render on the right hand side same as the filter/summarization. For me it would make for a better flow.

  12. Editor Icon - might be more intuitive to use a “pencil” or words…

  13. If you use the “Summarize” option e.g. count of rows and sum of quantity, then attempt to chart the results no charts render. While some of the chart options are greyed out they are still clickable. Should the chart visualizations be hidden when not appropriate for the data in question?

I appreciate your efforts!!

Error in preview2:

After clicking on a table on "explore data" I'm get this error with apache druid

Does this happen with all the tables in that Druid database? I’m having trouble reproducing this with our own Druid DBs.

in fact, this was happening with one table and I didn’t test with another, just this one, but now u mention it after I opened another table, the one that returned error works just fine

That is the magic of Maz. :sparkles:

4 Likes

Is anyone seeing CPU spikes and unresponsive UI?

Running Centos 7, Java 8.212, Metabase .33 preview 2

New design and builder are great by the way.

1 Like

It still glitchy. We build a report and narrow down to two columns visible, but CSV export for example still have all columns…

Also cant save an empty view. i.e. we have a view “stock greater on frontend than backend” its empty for the good times. Doesnt save - “value may be nil, or if non-nil, value must be an array of valid results column metadata maps.”

@maz
While testing it a bit further I came across that the binning doesn't work seem to anymore - at least not for latitude and longitude in the new query builder? - something that by the way doesn't work in the latest version for SQL either