Save details of your photos to a SQLite database and upload them to S3.
See Using SQL to find my best photo of a pelican according to Apple Photos for background information on this project.
These tools are a work-in-progress mechanism for taking full ownership of your photos. The core idea is to help implement the following:
- Every photo you have taken lives in a single, private Amazon S3 bucket
- You have a single SQLite database file which stores metadata about those photos - potentially pulled from multiple different places. This may include EXIF data, Apple Photos, the results of running machine learning APIs against photos and much more besides.
- You can then use Datasette to explore your own photos.
I'm a heavy user of Apple Photos so the initial releases of this tool will have a bias towards that, but ideally I would like a subset of these tools to be useful to people no matter which core photo solution they are using.
$ pip install dogsheep-photos
If you want to use S3 to store your photos, you will need to first create S3 credentials for a new, dedicated bucket.
You may find the s3-credentials tool useful for this.
Run this command and paste in your credentials. You will need three values: the name of your S3 bucket, your Access key ID and your Secret access key.
$ dogsheep-photos s3-auth
This will create a file called auth.json
in your current directory containing the required values. To save the file at a different path or filename, use the --auth=myauth.json
option.
Run this command to upload every photo in a specific directory to your S3 bucket:
$ dogsheep-photos upload photos.db \
~/Pictures/Photos\ Library.photoslibrary/original
The command will only upload photos that have not yet been uploaded, based on their sha256 hash.
photos.db
will be created with an uploads
table containing details of which files were uploaded.
To see what the command would do without uploading any files, use the --dry-run
option.
The sha256 hash of the photo contents will be used as the name of the file in the bucket, with an extension matching the type of file. This is an implementation of the Content addressable storage pattern.
The apple-photos
command imports metadata from your Apple Photos library.
$ photo-to-sqlite apple-photos photos.db
Imported metadata includes places, people, albums, quality scores and machine learning labels for the photo contents.
You can create a new, subset database of photos using the create-subset
command.
This is useful for creating a shareable SQLite database that only contains metadata for a selected set of photos.
Since photo metadata contains latitude and longitude you may not want to share a database that includes photos taken at your home address.
create-subset
takes three arguments: an existing database file created using the apple-photos
command, the name of the new, shareable database file you would like to create and a SQL query that returns the sha256
hash values of the photos you would like to include in that database.
For example, here's how to create a shareable database of just the photos that have been added to albums containing the word "Public":
$ dogsheep-photos create-subset \
photos.db \
public.db \
"select sha256 from apple_photos where albums like '%Public%'"
If you don't want to upload your photos to S3 but you still want to browse them using Datasette you can do so using the datasette-media plugin. This plugin adds the ability to serve images and other static files directly from disk, configured using a SQL query.
To use it, first install Datasette and the plugin:
$ pip install datasette datasette-media
If any of your photos are .HEIC
images taken by an iPhone you should also install the optional pyheif
dependency:
$ pip install pyheif
Now create a metadata.yaml
file configuring the plugin:
plugins:
datasette-media:
thumbnail:
sql: |-
select path as filepath, 200 as resize_height from apple_photos where uuid = :key
large:
sql: |-
select path as filepath, 1024 as resize_height from apple_photos where uuid = :key
This will configure two URL endpoints - one for 200 pixel high thumbnails and one for 1024 pixel high larger images.
Create your photos.db
database using the apple-photos
command, then run Datasette like this:
$ datasette -m metadata.yaml
Your photos will be served on URLs that look like this:
http://127.0.0.1:8001/-/media/thumbnail/F4469918-13F3-43D8-9EC1-734C0E6B60AD
http://127.0.0.1:8001/-/media/large/F4469918-13F3-43D8-9EC1-734C0E6B60AD
You can find the UUIDs for use in these URLs by running select uuid from photos_with_apple_metadata
.
If you are using datasette-media
to serve photos you can include images directly in Datasette query results using the datasette-json-html plugin.
Run pip install datasette-json-html
to install the plugin, then use queries like this to view your images:
select
json_object(
'img_src',
'/-/media/thumbnail/' || uuid
) as photo,
uuid,
date
from
apple_photos
order by
date desc
limit 10;
The photo
column returned by this query should render as image tags that display the correct images.
Datasette's custom pages feature lets you create custom pages for a Datasette instance by dropping HTML templates into a templates/pages
directory and then running Datasette using datasette --template-dir=templates/
.
You can combine that ability with the datasette-template-sql plugin to create custom template pages that directly display photos served by datasette-media
.
Install the plugin using pip install datasette-template-sql
.
Create a templates/pages
folder and add the following files:
recent-photos.html
<h1>Recent photos</h1>
<div>
{% for photo in sql("select * from apple_photos order by date desc limit 20") %}
<img src="/-/media/photo/{{ photo['uuid'] }}">
{% endfor %}
</div>
random-photos.html
<h1>Random photos</h1>
<div>
{% for photo in sql("with foo as (select * from apple_photos order by date desc limit 5000) select * from foo order by random() limit 20") %}
<img src="/-/media/photo/{{ photo['uuid'] }}">
{% endfor %}
</div>
Now run Datasette like this:
$ datasette photos.db -m metadata.yaml --template-dir=templates/
Visiting http://localhost:8001/recent-photos
will display 20 recent photos. Visiting http://localhost:8001/random-photos
will display 20 photos randomly selected from your 5,000 most recent.