Whether your data is hosted on the web, stored in a database, or saved in a local file, you can access it from Observable with these techniques.

Mike Bostock

Observable

Introduction to data

File attachments

Visnu Pitiyanuvath

Connecting to databases

Self-hosted database proxies

Annie Zhang

SQL cell

Wiltse Carpenter

Cloud files

Introduction to secrets

Cobus Theunissen

AWS S3 private file access

Justin Ting

Data table cell

Databases panel and schema viewer

Sample datasets

Ian Johnson

Curated datasets

SQLite

Custom database clients for use with Observable SQL cells.

Database Clients

Learn how Observable works, from first edit to publish.

Documentation

Getting Data In and Out

How does the storage of file attachment works? Are they hosted in your servers? And by which data protection terms does this hosting abide?

Thanks in advance!

Carlos Alberto Duarte Pinto

They are hosted on our servers. You can read more here: https://static.observablehq.com/assets/pdfs/Security+FAQ.pdf

It looks like this needs to be updated for the new paradigm without publish.

David Kirkby

Hi, I notice that File Attachments now support zip. It's awesome! No need to import Fil's @fil/jszip! I figure out how to use it, but I notice that there is no official documentation. It might be helpful for other users to find some references here. I'd be happy to contribute if that's the case!

Radamés Ajna

hi @fil thanks! The stlib with zip is very convenient!

.zip() is now documented at https://observablehq.com/@observablehq/fileattachment-zip ; enjoy!

Can we please make FileAttachment accept a formatted string or a string variable as input? It is very inconvenient having to hard code each file name that we want to use.

Ameya Patil

This restriction is needed to perform static analysis of notebooks to improve security by knowing which file attachments are in use. If you want to reference file attachments dynamically, you can create an array of them: files = [FileAttachment("1.txt"), FileAttachment("2.txt"), …] and then files[0], files[1], etc.

Makes sense! Thank you for the suggestion

Noteworthy here too that you may pass in size parameters: e.g., `FileAttachment(...).image({ width: 400 })`

Aaron Kyle Dennis

How can I replace a file (file.csv) with a newer file by the same name (file.csv)? I don't want a "file@1.csv" for the new filename.

daktari

You can’t; files are immutable, so the replacement file must have a new name. That said, you can name your file like “file-20201101.csv” before uploading if you’d prefer to a different unique rather than the auto-generated “file@1.csv”.

I understand but I need to keep the file name unchanged. Since I only want replace the content of the file and not touch any code line. To have at least a workaround I fork my notebook, delete the file, fork again and upload the new file.

If you use the “replace file” feature, all references to the old file will be replaced with references to the new file automatically.

What happens to file attachments when I export? I tried to export a notebook with attached images, but the files did not show up in the resulting tar file and the notebook broke...

Andrew Wooldridge

It seems like 7zip on Windows struggles with our tar.gz files; see Andrew's forum thread here: https://talk.observablehq.com/t/exporting-with-file-attachments/5094. We'll look into whether that's a bug we can fix or if we should offer a .zip download too.

Toph Tucker

Would be nice to add links to FileAttachment API and examples from here.
Every time I use Google Sheet, I search for API docs and land here, but there is no references to https://observablehq.com/@observablehq/xlsx

Ilyá Belsky

Agreed, we should definitely add the .xlsx() function here also.

What is `well-typed CSV` ? Is there a spec that describes how to specify CSV column types?

Anatoli Babenia

Defined here: https://github.com/d3/d3-dsv/blob/main/README.md#autoType

It still doesn't explain what a well-typed CSV is.

For me `.csv({autotype: true})` would be more readable than `.csv({typed: true})`, because there is no such thing as a typed CSV. It might be a bit too late to rename it though.