Preview upcoming changes to the GUI query builder (updated)

@brianwilson Just a couple of notes :slight_smile:

  1. Changing date format has been possible since version 0.30 I think
  2. All documentation links will fail, since it uses the version number in the link - and since this is a special release, then there’s no documentation page for that. But maybe we should have a redirect on the server to latest instead of 404.

Is there any way that Metabase may support output the image?

Like the result of the mysql is kind of “http://www.xxxx.jpg” then, it can show like the picture, in that case, is very easy for the people who may don’t know much about HTML or PHP, but can still do progaming.

Tks
Scug

@ScugZd That’s already supported, either via table column formatting (only thumbnail) or via text cards. Are you looking for something different?

Tks Flamber, I tried, the problem I have now is:

  1. If I use the “Browse Data” model, yes, I can change it to image formate, and like you said, only thumbnail, too small, may we get it bigger?
  2. If I “write SQL”, it seems the “formate” option won’t show up, which means when I “click on the column heading”, nothing happen.

@ScugZd

  1. Currently no, but there’s a request to allow bigger images - go and upvote by clicking :+1: on the first post.
  2. Click Visualization settings :gear: (to the left, just below the visualization - on the older versions, just above), since SQL queries doesn’t have drill-thru menus.

If you have other questions, then it’s probably best to start a new thread unless it’s directly related to the new query builder.

1 Like

I finally had a chance to take it for a short spin - just using the demo data. Oh boy do I like it!

Just one cosmetic glitch I really noticed so far. I tested with Firefox - and there are a number of places where unnecessary scroll bars are shown. It didn’t come as a surprise though as that has also happened in various previous versions - so almost not even a regression. :wink:

Overall excited to see this and what what this big step can do for Metabase and its users! Congrats :tada:

3 Likes

Can you open a separate thread to discuss images in result renderings?

We’d like to keep this thread focused on direct feedback regarding the new version of the query builder

Tks so much Flamber!

Overall it looks great, I think lots of users will be glad to be able to smush together tables where it’s needed but where they aren’t comfortable with SQL. The comments so far have covered things I was going to say, except for 2 suggestions:

  1. seeing as joining tables can introduce quite a lot of columns to the table visualisation, it would be useful to have a “deselect all columns” type of button (i.e. next to the “Reset to default”) so a user can just pluck out the handful of columns they’re interested in, rather than clicking away 10’s of uninteresting columns one-by-one.

  2. not sure if it’s a recent addition, but the “preview” slider in SQL questions is confusing. I would prefer if it defaulted to “off” since if I know I’m going to get loads of rows I’ll add a “limit 10” or appropriately aggregate in the query. On the other hand I can see it might be a decent foot-gun-guard for people new to SQL - in which case could it go on top left next to the database selection drop-down, so it’s close by and obvious to turn off? Or even as a profile-level setting like “Thanks but don’t default to previewing SQL please”.

Thanks for trying out the preview, Lindsay!

  1. Yeah, I agree that it’s a drag right now to deselect all columns in a big table in order to then only add back the ones you want. Also agree a deselect/select all would be a welcome addition.
  2. Yep, we don’t feel that we got this feature right, so for the time being we’ve completely removed it, and plan on revisiting it or something like it based on feedback.

A question for all of you: have you or any of your less technical users tried out the simpler summarizing or filtering tools available from the default view of a table or when viewing a GUI-based saved question? (I.e. the Summarize and Filter buttons below the title of the question.)

Any thoughts on how those interactions feel as a simpler, alternative way to filter, aggregate and visualize data? Thanks for your thoughts!

@maz

I’ve only had a single non-tech user test the preview, but the “Browse data” is not a great entrance, when you have a lot of tables (200+), so I showed her alternative router via “New custom question”. But that takes you directly to the editor.

I then explained the new join function, which she really liked - since she can do much more by herself instead of having me make SQL/view.
She was a little confused when the buttons (join etc) would show/hide or had color or change position (as mentioned by others).

Really liked the new visualizer selector - but didn’t understand the viz/raw toggle until I explained it.

Took a little while to understand that metric and group by was hidden under “Edit summary”, while the filter was displayed - while everything was shown in the editor.
And sometimes she couldn’t group by month in summary side panel, while she had that functionality if she did everything in the editor. I noticed that this depended on how you start creating the question (browser/custom).

Had her make a couple of the existing advanced questions used in some dashboards - came back 20min later - she made all of them, but she chose the editor every time instead of summarize side panel.

Thanks @flamber!

the “Browse data” is not a great entrance when you have a lot of tables (200+)

Yeah, in the future we're looking at things like adding filtering/search, and/or the ability to pin useful tables or custom datasets to the top of the list. My assumption based on what you wrote is that what makes it "not great" is that it's hard to find the table you're looking for, or hard to know which tables might be interesting/useful. Is that right?

Really liked the new visualizer selector - but didn’t understand the viz/raw toggle until I explained it.

The position and styling of these elements in the preview seems to have confused lots of people, so we've tweaked the positioning of that toggle and are working on a different visual treatment for the visualization button.

Took a little while to understand that metric and group by was hidden under “Edit summary”, while the filter was displayed - while everything was shown in the editor.

Was there a different label she expected rather than "edit summary?" (After we released the preview build, we've since changed that to just always say "Summarize" FYI.)

And sometimes she couldn’t group by month in summary side panel, while she had that functionality if she did everything in the editor. I noticed that this depended on how you start creating the question (browser/custom).

I'll have to see if I can repro that.

Had her make a couple of the existing advanced questions used in some dashboards - came back 20min later - she made all of them, but she chose the editor every time instead of summarize side panel.

For this scenario, I think this is "working as intended." I.e. if you're in a mode of creating specific things, which might be somewhat complex/advanced, that is the flow we intend for you to use. (Whereas the flow starting from Browse Data, or starting from an existing chart/question is meant to be the path of lightweight exploration or modification.)

@maz

...it’s hard to find the table you’re looking for...

Yes and no. All the tables are interesting :smile:, but like most other structures, some are only really interesting, when joined with other tables. It's just much faster to have a simple table searcher, when you have hundreds of tables. Alternatively you could to use the browsers built-in page search.

The position and styling of these elements in the preview seems to have confused lots of people...

I'm following the PR and do test-builds, so I really like the progress, but I showed her the preview release last week.

Was there a different label she expected rather than “edit summary?”...

No, I think she was just used to be able to see everything - like the old QB.

I’ll have to see if I can repro that.

I haven't tested the latest notebook-mode, but watched her show it several times. It seems like using Custom creates a different "Summarize by" row, when created from the editor, than when going thru Browse and just using the Summarize-button.

For this scenario, I think this is “working as intended.”...

I agree, meaning I understand your point, but I just wanted to show how a non-tech user mis/understood the Summarize vs Editor.
It might have something to do with habit from QB - or maybe the issue she encountered above (Browse vs Custom).

I hope we're going to get at least one more preview/candidate, just because it's such a big change for the user.

@flamber, thanks for the reply.

It’s just much faster to have a simple table searcher, when you have hundreds of tables.

What do you think is contributing to the feeling of the browser being slower than the "simple table searcher"?

  1. It doesn't have a way to filter or search the list of tables.
  2. The arrangement of the tables on the page (grid layout instead of a vertical list)
  3. The size of things on the page (large buttons instead of a compact list)
  4. Something else?

@maz
When there’s 200+ tables, then I don’t think 2) and 3) would help, but having a 1) search/filter makes the difference. That’s what makes QB great - even with many tables/columns.
But perhaps some users should just use Custom, while others should use Browse.

I’ve never used /browse or /reference until this preview, so maybe I’m biased, but when you have a lot of tables or columns, then it becomes difficult to use without a search/filter.
Admin > Data Model has a table filter, which works great, but lacks a column filter (some tables has 130+ columns).

I really don’t know if it should be a priority for 0.33 - perhaps most people don’t have hundreds of tables/columns. I would prefer to hear from other people in the forum.

2 Likes

I also dislike “browse” for more than, say, 15 tables. It’s harder to scan a 2D grid than a 1D list. Search might help.

Thought you all would like to know we just released an updated second preview. :slight_smile: https://github.com/metabase/metabase/releases/tag/v0.33.0-preview2

Thanks again for all your feedback and thoughts!

2 Likes

Hi there,
After having tested both versions, I really appreciate the changes you have made to the joins since the last preview version :+1:
I would love to see a few more modifications made to the custom question interface.

  1. It is a hassle to load in all the columns first, just to afterwards deselect all but 3 columns. Say you need to join 5 tables and each might vary from 10 to 50 columns - this becomes very time consuming and not that great for performance either. - Hope you guys can do your magic here.
  2. As this is a huge improvement for our none tech people, there are still cases where you would need to convert it to SQL, for example you can’t say that COLUMN A = COLUMN B
  3. Would be really cool if the add column would allow for functions like CONCAT(), DATE_FORMAT() and other SQL functions inside a select statement - and not only limited to calculated values

Again: Love the changes you guys have already made - Huge improvement IMO

1 Like

I tried out preview2, it looks more polished for sure. I have 1 bug and 4 comments:

  1. bug: when joining at least one table, the table visualisation settings pane is broken, and doesn’t show up at all. The console is repeatedly showing an error “TypeError: t.field is undefined app-main.bundle.js:5:1112115”. It doesn’t do this when there is only one table (no joins). Tested with Firefox Quantum 67.0.3, this wasn’t happening in preview1.

  2. The alerts button (bottom right) could do with a hover tip to say what it’s for, like how the downwards arrow says “Download full results”.

  3. The way the filter pill-shaped buttons flow / wrap in the editor is really nice, they stay in their little lilac zone which expands as necessary, and each filter’s text stays inside the button and all of the buttons are visible. I think it should work the same in the visualisation area, where instead they don’t wrap and the text overflows the buttons, and it only takes 2 or 3 wordy filters to start overflowing.

  4. Choosing summarisation groupings should probably work the same in the editor and the pop-out pane. That is, clicking a column name should add it to the list of grouping fields, and clicking the “x” removes it. At the moment it works like that in the editor, but in the pane when I click a column name it de-selects all other grouping columns (instead I should click the “+” to the right of the name to add). I think the editor’s way is more intuitive but whatever way is chosen I think it should be the same sort of interaction.

  5. About choosing columns, I prefer the older accordion style list where you can search or click a table name to expand the columns, because the data warehouses I work with often have lots of related tables (range is about 1 to 10). So coming from that perspective the flat style list in the visualisations area feels like it needs to make a bigger deal out of the table names and where each table starts/ends. Like increasing the table name font size or a subtle border line between tables, to help break up the list when scanning through columns (with or without filtering). Similar to above, it probably should be the same type of UI in both places - i.e. in the visualisation area there’s a flat style list for filters and summaries, but in the editor the filter, grouping, and join columns choosers uses the accordion style list.