More feedback • In the pre release I am missing the ability to nest pages more than one level, any deeper hierarchies crashes the page because the left sidebar menu doesn´t seem to support it • I think I've mentioned it, but the ability to search, like in mdBook and similar • A good example of both a nice clean sidebar + search integrated is the Tailscale documentation

+1 for search option

Released Changes • The way we interact with data under the hood has changed; there shouldn't be any noticeable impact yet, but if you encounter odd behavior please let us know • Error and Loading states have been improved If you are using the

usql

versions in

package.json

, you can run

npm update

to update to the latest versions, and

npm i

to install them Happy sqling!

Hi @bland-thailand-4711

• Error and Loading states have been improved

Does this relate to this case, where the Datatable or Big Value become red for split seconds before the data is loaded? If yes, then how did you solve this? The funny thing I have just recently told Adam I didn't need this issue anymore, since I'm not using USQL, but today this issue cropped up in my case too. Now I'm using the standard Sveltekit's way to work with duckdb through

+page.server.ts

and I managed to find a way to pass dynamic parameters to the load() function through a combination of

url.searchParams

, `invalidate`and

depends

functionality to re-trigger the

load()

reload when parameters change, but now it leads the UI component (DataTable) to fully reload and flash the red empty state. So I wonder how you solved it and if I can adapt it to my case.(second gif). Thanks

How do I use csv/tsv files with empty/null cells? I get an exception when trying to read a csv where some cell is missing values:

Unsupported object type: null

I've tried different quotations etc.

Check out the

IFNULL()

function form duckdb. https://duckdb.org/docs/sql/functions/utility.html You could add it the query where you use it and deal with the

null

before the JS layer.

Can I hide a page from being added to the sidebar? Using frontmatter or some file convention. I want a page to be directly linked only.

Released Changes •

build:sources

now uses cursors / streaming for supported databases, this should make the command faster, and able to support arbitrarily large datasets (the client / network limits still apply though, so be sure to filter your data when you put it on the page!) • Deploy screen shows the correct environment variables to deploy your project. The next release will have configuration of the sources in the settings panel (instead of needing to do the yaml files yourself) • Breaking: Environment variables now use

__

instead of

_

to separate levels, so

EVIDENCE_SOURCE_mysource_some_option

needs to become

EVIDENCE_SOURCE__mysource__some__option

. Refer to the new deploy section if you are unsure, those will be correct for the new system •

build:sources

has some other ergonomic improvements •

Value

and

BigValue

now use loading state as well If you are using the

usql

versions in

package.json

, you can run

npm update

to update to the latest versions, and

npm i

to install them Happy sqling!

Big fan of the usql release! Have a question around interactivity - specifically for allowing users to append new source data from a component, and have relevant dependent queries/components refresh. This would be useful in the context of reviewing / modifying classifications (from GPT, for example), or for setting scenarios for a forecast (we do the latter at trafficprojection.com, and are integrating the former atm). What I imagine would be required to make that possible: 1. Append records to a parquet source from a grid-like component (lots of existing grids that can be used). 2. Refresh dependent queries as source data is updated (similar to modifying an input today). Is that a) possible on this branch, b) something on your radar for future, or c) incompatible with the static build approach?

Hi, after bumping evidence version to

20.0.0-usql.29

I'm getting:

Parquet error: Incorrect number of rows, expected 7832 != 208 rows

when building postgres data source. Here is my plugins list:

"@evidence-dev/bigquery": "^2.0.0-usql.3",
    "@evidence-dev/core-components": "usql",
    "@evidence-dev/csv": "^1.0.0-usql.4",
    "@evidence-dev/evidence": "20.0.0-usql.29",
    "@evidence-dev/faker-datasource": "usql",
    "@evidence-dev/mysql": "^1.0.0-usql.4",
    "@evidence-dev/postgres": "^1.0.0-usql.3",
    "@evidence-dev/sqlite": "^2.0.0-usql.2",
    "@evidence-dev/usql-datatable": "^1.0.2",
    "mermaid": "^10.4.0"

Released Changes • Toast notifications got a facelift • New

Dropdown

component ◦ This is the first of several input components that are DuckDB backed •

debounce

has been applied to query stores ◦ If a query changes rapidly (e.g. scrolling through a dropdown), it won't fire the updated version until changes stop for 500ms • Performance Guardrails have been added ◦ Toast / console notifications will let you know if the data volumes you are using are growing beyond a certain point. ◦ These aren't hard rules, just reminders to make sure you are only loading what you need • A fix for parameterized page pre-rendering has also been added If you are using the

usql

versions in

package.json

, you can run

npm update

to update to the latest versions, and

npm i

to install them Happy sqling!

Hi! Coming back to my project after a few weeks and what I think was working before doesn’t work anymore. I am using the latest version of

next

I’d like to query some

parquet

files that I have already created in advance. The code that I think worked a few weeks ago was:

```listings_details
select 
    *, 
    latitude::string as lat,
    longitude::string as lon,
from '/abcd/fct_listings_light.parquet' 
where id = ${$page.params.listing}
```

But now, I get a

ran successfully but no data was returned

. Is there some specific config under

sources

to put in place to be able to query those parquet files?

Howdy, how can I create a link to a headline in evidence?

# Headline A

Checkout what we are saying [under C](#Headline C)

## Headline B

## Headline C

• The build breaks in prod mode due to missing links • We get numbers after our headlines that only show after build, like:

#Headline C5

. • Scrolling down to the headline doesn´t seem to work Is there a best practice or documentation?

Hi, I'm having another issue with usql release https://github.com/evidence-dev/evidence/issues/1389

What is the current correct combination of tags for "latest usql"? I changed

"@evidence-dev/bigquery": "^2.0.0-usql.1",

to

"@evidence-dev/bigquery": "^2.0.0-usql.4",

and now the bigquery connector won´t load bus instead gives an error:

[!] "undefined" could not be loaded as a plugin, it is missing a name or main field.

Here are my deps

"@evidence-dev/bigquery": "^2.0.0-usql.4",
    "@evidence-dev/core-components": "usql",
    "@evidence-dev/csv": "^1.0.0-usql.4",
    "@evidence-dev/evidence": "usql",
    "@evidence-dev/faker-datasource": "usql",
    "@evidence-dev/mysql": "^1.0.0-usql.1",
    "@evidence-dev/postgres": "^1.0.0-usql.1",
    "@evidence-dev/sqlite": "^2.0.0-usql.2",
    "@evidence-dev/usql-datatable": "^1.0.2",

👋 Hey everybody, For those of you using the prerelease; there are a few breaking changes on the latest version ( you can ignore this if you haven't updated to the latest pre-release yet ). You'll need to: • Update `evidence.plugins.yaml`'s

databases

field to

datasources

instead • Update any

connection.options.yaml

files to have their values base64 encoded. Alternatively; you can update those values in the settings UI and it will be automatically formatted as needed

Is there a "how to get started with next/usql" gist? I'm trying to piece together via slack convos but it seems to have some issues that I can't find an Issue or thread about. mostly this, but also in general what should be in my deps

[!] /Users/dpage/Projects/Data-Platform/frontend/sources/reports is not a valid source; skipping
ENOENT: no such file or directory, open '/Users/dpage/Projects/Data-Platform/frontend/sources/reports/connection.yaml'

has renamed the channel from "pre-release" to "migration"

With the recent release of Universal SQL, going forward, this channel will be used to aid those migrating legacy Evidence projects to USQL (>v24) • Migration guide should cover most requirements • We expect this won't cover everything, so if you are having trouble, we're here to help!

set the channel description: Tips and Help on Migrating to Universal SQL

set the channel description: Tips and Help on Migrating Legacy projects (<=v23) to Universal SQL (v24+)

well now i have a good project for tomorrow. excited to check it out guys!!

Alright a very specific migration question related to filters

is there a way to access the read_csv_auto params when using CSVs in usql?

Hi all, how do I configure Motherduck with the latest version? I was using the connection.yaml

name: mydb
type: duckdb
options:
  filename: md:?motherduck_token=<token>

it doesn't work. Seems it still expecting duckdb file Slack Conversation

finally messing around with usql. After going through the migration steps, I'm getting

500:TextInput is not defined

when trying

<TextInput
    name=name_of_input
/>

I am able to use the dropdown input as expected though. I've run

npm install @evidence-dev/evidence@latest @evidence-dev/core-components@latest

to ensure I am on the latest version. Hoping I'm missing something obvious

In verison 24, strings are converted to timestamps when displayed in DataTable, and timezone is also converted. Is there any way to disable this? I want to display the string YYYYY-MM as it is. If the timezone is Asia/Tokyo, 2024-01 will be displayed as 2023-12-31.

The Pie Chart example in the document does not work with evidence v24. It worked with previous version except quotation mark issue. https://docs.evidence.dev/components/custom-echarts#pie-chart It may be all ECharts. 1. Use single quotation to select literal wrong

select "Apple" as pie, 60 as count
union all
select "Blueberry" as pie, 70 as count
union all
select "Cherry" as pie, 40 as count
union all
select "Pecan" as pie, 35 as count

ok

select 'Apple' as pie, 60 as count
union all
select 'Blueberry' as pie, 70 as count
union all
select 'Cherry' as pie, 40 as count
union all
select 'Pecan' as pie, 35 as count

2. The query name in the data parameter does not work.

<ECharts config={
    {
        tooltip: {
            formatter: '{b}: {c} ({d}%)'
        },
        series: [
        {
          type: 'pie',
          data: pie_data,
        }
      ]
      }
    }
/>

data: [...pie_data]

or

data: pie_data.slice()

worked. I don’t know why. I’m not familiar with Javascript.

This seems like an obtuse question, but I am stuck. I upgraded Evidence to v24. Now need to migrate the data source and possibly SQL. However, when I go to settings the drop down for Data Sources/Datasource Type is empty. Moreover, ‘npm run sources’ yields a missing script error. How can I set up a data source to MotherDuck?

Hi Evidence team! First, thanks for all the great work on Evidence -- it's truly a differentiated product out there in the BI space! I'm Director of Data for a PE firm that works with a lot of smaller, not-necessarily-tech-focused companies, and evaluating Evidence as a BI tool to surface KPIs to executives in those companies. I have a couple questions about the latest USQL version: • Is there an option or flag somewhere to build the legacy way and sacrifice the interactive filters? • I looked in the network requests that evidence makes, and it looked to me that it GETs the parquet files that it needs when the filters are triggered -- is that right? And, if so, am I right in thinking that we'd have to keep the parquets to a reasonable size? • What is the "recommended story" for authorizing certain reports to certain users? With legacy, I could envision a story around authenticating GET requests for given URLs, but it seems more complicated in USQL because evidence now makes some requests for the parquet assets.