Hey Folks, I'm interested in extending the DataTable so that we can specify default sorting behaviour when grouped (section/accordion). Changes to the current default behaviour: • Currently, grouped DataTables defaults to sorting by the first column. I would like to have say a property

sortByCol

• Within the groupings it would be great to also be able to specify the sort order. Perhaps a comma separated list would work well e.g.

sortByCol=A,B,C

where the groups will be sorted by

A

and rows within each group sorted by

B then C

Can evidence do write-back?

It would be really cool if <DataTable> could have a "connectGroup" property to sync highlighting to adjacent charts 😇

Can I switch the map provider somehow? From leaflet to maplibre

Just want to resurrect this thread as it's a feature I am also interested in - exposing the currently selected row. It's ok to create a github issue for it? https://evidencedev.slack.com/archives/C023MMFSWKU/p1715592585413349

one thing i find myself wanting more and more is either of • the ability to have multiple input components that control the same variable • a floating component that remains in the viewport, when the page is scrolled beyond the original "location" of the input component in some of my current use cases, it's common for me to want a single input variable to affect all charts (or a collection of them) on a page - typically to effect a timeshift . i find this amusing because i used to desperately want the opposite when I worked in Grafana - the ability to have per panel variables. the workaround with Evidence is easy - I can just create another on-page query and a "fresh" dropdown that targets only that query (the caveat being there are more queries to update and manage). and i can still use "connectGroup" to link all the panels so they respond alike to cursor positioning. in reality, both of these options might be useful, for different use cases. there's definitely some occasions when the floating component won't be great in terms of responsive layout. i'm curious if anyone else has wanted to do this, or if i'm maybe trying to put too much on one page. i'll be publishing my latest work shortly so will update with a link, in case anyone has suggestions for how to do it differently

Does anyone know if there’s a tweak to use a TSV instead of a CSV as source?

Having different number format for tooltips and Axis labels would be nice. For example by adding a yAxisFmt property

🔥 or 👎🏼

I have user that needs to be able to either 'Search' within an AreaMap component. That search would be across the geoID - and once selected they want to effectively 'select' the associated polygon on the map (like they zoomed to and clicked on it). The reason for this is they don't know geospatially where the data they are interested in is, but they do know the geoID.

Nice to have … Also working with SuperSet they show the number of records by default in the search box. Looks nice and my end-users find it usefull. Looks like this…

I love the

Grid

component to help organize pages and better utilize space on a page. However I sometimes find myself wishing that I could adjust the widths of each column. For example, if I have a

BigValue

component on the left and a

LineChart

on the right in a 2-column grid, the BigValue doesn't take up (or need) nearly as much space as the line chart. Wondering if a prop to set column widths would be possible? E.g.

<Grid cols=2 widths={[0.3, 0.7]}>

I am not sure if this is a `feature-idea`or I just didn’t find the right configuration. I was expecting the

Data Table

in

explore/console

to be clever enough to know which rows I am seeing and just loading those rows. If I have 10M rows, but I am displaying only 100, why do I need to load all of them and wait several seconds?

I know this sounds strange as there is no server and this is the kind of thing you do with a server. But AFAIK duckdb is very clever not to load everything when it is not needed… so why not implement in client side a logic that is common on server side?

Hi <!channel> We are moving over #feature-ideas to GitHub Discussions We love getting ideas from the community, but it's easy for slack messages to get lost, and hard to see if anyone has asked for the same feature before. With discussions, you can: • See if the same question has been asked before • Upvote ideas you want supported • See which responses have been marked as answers, StackOverflow style For more details on the move, see this discussion.

Hello everyone, I love the SqlConsole component. I think it would be a lovely idea if we could edit the input message and change the name of the component to something else than SQL Console. Thanks

Hey all - is there some way to define and expose data definitions to an end user (like Looker's ℹ️ icon that showed sql and user defined description)? I find that my sql queries often categorize dimensions, either to label data for users with case statements or to filter charts with where clauses. For example,

when column_a in ('x', 'y', 'z') then 'label_i' else 'label_j' end

. Easy to do in sql but then each label or chart needs to be explained to non-sql users in some way i.e. user doesn't know what label_i and label_j are made up of (and same goes for me as the project maintainer...I barely remember what I did yesterday let alone a column label I defined several months ago). I could copy/paste the definition into the markdown but it will be out of date as soon as the sql code is changed. So I find myself wanting something like dare I say a LookML-like semantic layer...or at the very least the ability to write a piece of logic and reuse it like a dbt macro i.e. define a 'data_labeler' in a way that can write the sql and also the markdown using the same piece of reusable code like a variable. Then the sql and the markdown will always be linked. One could define this stuff in the upstream transformation layer(s) but then the same problem of explaining what goes into each column or chart to the user remains. Here it would make sense to have clear and accessible data documentation to the end user e.g. dbt docs but I think it would be quite powerful to have this embedded right into the Evidence product so that a user can view and interpret data all in one place (added bonus: less likely they have to ask any data team members either 👀) ^^ alternatively, anyone know of any current implementations for something like this?? Would love to hear how Evidence can be configured to replicate this!

It'd be great to be able to hide certain pages using environment variables. I love that you programmatically render certain parts of a page using environment variables, and it'd be even better to be able to do this at the page level.

The evidence looks like a webpage or like a report that you can export easily to PDF. And I like it! But time to time, I'm feeling that I would like to build more like a "dashboard" with grid layout. Have you considered (or even someone implemented) something like this with custom layouts? Attached here example screenshot that I quickly found with Google (I have not done it).

Hi I've been trying to figure out how I can add a custom component to evidence. I saw the simple examples in the

evidence-labs

repo. But was hoping for some more detailed guidelines, especially on extending the

DataTables

component & on how to enable cross filtering of visuals based on selections in one of them. Regards

DataTables

- I would really like row-wise actions/events e.g. a checkbox per row & a button at the bottom which could trigger some arbitrary function in the backend via. some post request containing the json of selected rows? Would such a thing be possible to mod evidence for? Regards cross-filtering I want to be able to select on a summary view of information, say in a bar-chart & to then be shown the raw-records from the underlying data that went into creating the summary view. How would such an event/callback style action be possible, how would I modify or design a component for it?

Hey, a feature idea (and perhaps simultaneously checking if anything already exists here) At my company we are a dbt house. We have a centralized dbt project that builds our entire DWH. We leverage exposures to ensure we can keep track of downstream impacts when working with our dbt models. My feature idea is more of a problem than a solution: I think it would be very valuable to have some procedure to automatically generate exposure entries for evidence pages. Basically, some approach that would read an evidence project, figure out metadata on the pages and what dbt models they rely on, and then modify the dbt project exposures to list all pages and map the upstream dependencies they hold towards dbt models. Happy to discuss if other people have similar needs, perhaps if someone is already doing/building something along these lines.

Hey team, It would be great to have a native way to add a description/definition to a chart. This could be implemented with a hover effect or an information icon in the corner of the graph. Here’s how we’ve handled it on our side for now, but having a native solution would be much easier to maintain!

Has anyone tried Bruin https://bruin-data.github.io/bruin/ ? So far, the evidence sources has been good enough for my use case. But now it start to become more complex and I started to look for simple solution for this. I'm not experienced data engineer so I'm looking this with fresh eyes 🙂 I found this quite interesting option for ETL pipeline that could build the source data for Evidence. Would love to hear your opinions on Bruin and ideas how to integrate it with Evidence.

I would like to have BigValue but also include the target, for example show value 30/50 (50 is the target) or some other way. Has anyone done something like this? Any ideas how to implement?

Hi, dropping some ideas here. Kind of suggestion/kind of checking if someone might have built/seen something similar already. I've made the effort to search for this previously but have found nothing. I'm assessing evidence and one of the things that feels promising and has me excited is that, with all the dashboards, visuals, queries, etc defined as code, I can parse that and roll integrations with other tools. The most obvious one to me is dbt, which we rely on at my team. For example, I can think of: • in our dbt project we have a

reporting

schema which is the "public" layer of models which data products should rely on. I can see how I could create a

select * from table

.sql

file for each table in my reporting schema and just add it my source folder. • The other way around: I could parse the evidence project to introspect what are all the models that are being used in each page, and then create an

exposure

in the dbt project to map the dependency properly so that we can take it into account when modelling in dbt. So I guess my questions are: • Are there any tools like this already out there? • Does someone have experience trying out this kind of stuff? Thanks

Hi guys, I’ve noticed that cross-filtering on data tables is in the roadmap, is there any ETA on when we can expect that? Are there any workarounds to achieve similar functionality in the short term?

Hi guys, it would be great to support query functions in the

#each

clause in a loop. I was hoping to do something like the following, but it doesn’t seem that. itwill work without two seperate queries:

<Tab label="Best Performers">
    <div class="tabtextcontainer">
        {#each ranked_with_names.limit(5) as best}

        <Value data={best} column=deviceName emptySet=warn
               emptyMessage=""/>

        <div class="tab_percent_container">
            <Value data={best} column=current_utilisation fmt=pct1
                   emptySet=warn emptyMessage=""/>
        </div>

        <div class="tab_delta_container">
            <Delta data={best} column=percentage_change fmt=pct1
                   symbolPosition=left emptySet=warn emptyMessage=""/>
        </div>

        {/each}
    </div>
</Tab>


<Tab label="Worst Performers">
    <div class="tabtextcontainer">
        {#each ranked_with_names.reverse().limit(5) as worst}

        <Value data={worst} column=deviceName emptySet=warn
               emptyMessage=""/>

        <div class="tab_percent_container">
            <Value data={worst} column=current_utilisation fmt=pct1
                   emptySet=warn emptyMessage=""/>
        </div>

        <div class="tab_delta_container">
            <Delta data={worst} column=percentage_change fmt=pct1
                   symbolPosition=left emptySet=warn emptyMessage=""/>
        </div>

        {/each}
    </div>
</Tab>

Would alerting be in the pipeline - send slack alert once a metric reaches a threshold? Or perhaps emulating the same behaviour using an evidence style-guided way for writing metric tests (they will run an assertion on committing and then raise an alert or message from github-action etc. if failing assertion)

Hi friends 👋, I'd be interested to learn more about folks' thoughts on potential ways server sent events might be used to push data updates to Evidence apps. Not sure of feasibility, but I'm imagining potential configuration options for an SSE source could be noted in a connection.yaml file, that would then expect something like a FastAPI/Starlette StreamingResponse, which could then (if provided) be accessed in the Svelte components through a Writable store (or source, with sveltekit-sse). Or if not SSE, perhaps the solution might be something more like streaming Arrow record batches to duckdb-wasm via insertArrowFromIPCStream? I'm new to these tools, and any feedback or insight on what/where to further explore would be appreciated - thanks!

Would it be possible to add a shape identity channel (Marks and Channels (Ch 5)) to

PointMap

? It would be great to use it as a categorical identifier in maps. Especially when color hue is already in use (currently the only categorical identifier in

PointMap

)