I'm starting this issue without a clear reproduction in case someone else has seen this behavior, and to use the issue as a notebook for research.

I'm using Datasette with the SWITRS data set, which is a California Highway Patrol collection of traffic incident data from the past decade or so. I receive data from them in CSV and want to work with it in Datasette, then export it to CSV for mapping in Felt.com.

Their data makes extensive use of codes for incident column data (1 for Monday and so on), some of it integer codes and some of it letter/text codes. The text codes are sometimes blank or -. During import, I'm creating lookup tables for foreign key references to make the Datasette UI presentation of the data easier to read.

If I import the data and set up the integer foreign keys, everything works fine, but if I set up the text foreign keys, CSV export starts to fail.

The foreign key configuration is as follows:

```

Some tables use integer ids, like sensible tables do. Let's import them first

since we favor them.

for TABLE in DAY_OF_WEEK CHP_SHIFT POPULATION SPECIAL_COND BEAT_TYPE COLLISION_SEVERITY do sqlite-utils create-table records.db $TABLE id integer name text --pk=id sqlite-utils insert records.db $TABLE lookup-tables/$TABLE.csv --csv sqlite-utils add-foreign-key records.db collisions $TABLE $TABLE id sqlite-utils create-index records.db collisions $TABLE done

Other tables use letter keys, like they were raised by WOLVES. Let's put them

at the end of the import queue.

for TABLE in WEATHER_1 WEATHER_2 LOCATION_TYPE RAMP_INTERSECTION SIDE_OF_HWY \ PRIMARY_COLL_FACTOR PCF_CODE_OF_VIOL PCF_VIOL_CATEGORY TYPE_OF_COLLISION MVIW \ PED_ACTION ROAD_SURFACE ROAD_COND_1 ROAD_COND_2 LIGHTING CONTROL_DEVICE \ STWD_VEHTYPE_AT_FAULT CHP_VEHTYPE_AT_FAULT PRIMARY_RAMP SECONDARY_RAMP do sqlite-utils create-table records.db $TABLE key text name text --pk=key sqlite-utils insert records.db $TABLE lookup-tables/$TABLE.csv --csv sqlite-utils add-foreign-key records.db collisions $TABLE $TABLE key sqlite-utils create-index records.db collisions $TABLE done ```

You can see the full code and import script here: https://github.com/radical-bike-lobby/switrs-db

If I run this code and then hit the CSV export link in the Datasette interface (the simple link or the "advanced" dialog), export fails after a small number of CSV rows are written. I am not seeing any detailed error messages but this appears in the logging output:

``` INFO: 127.0.0.1:57885 - "GET /records/collisions.csv?_facet=PRIMARY_RD&PRIMARY_RD=ASHBY+AV&_labels=on&_size=max HTTP/1.1" 200 OK Caught this error:

```

(No other output follows error: other than a blank line.)

I've stared at the rows directly after the error occurs and can't yet see what is causing the problem. I'm going to set up a development environment and see if I get any more detailed error output, and then stare more at some problematic lines to see if I can get a simple reproduction.

At the bottom of database pages, datasette gives users the opportunity to download the underlying sqlite database. It would be great if that could be served gzip compressed.

this is similar to #1213, but for me, i don't need datasette to compress html and json because my CDN layer does it for me, however, cloudflare at least, will not compress a mimetype of "application"

(see list of mimetype: https://developers.cloudflare.com/speed/optimization/content/brotli/content-compression/)

I have a schema that uses numbers for a column (actually it's a boolean 1 or 0 but SQLite doesn't have Boolean). I can't seem to get the facet to work or even filtering on this column.

My guess is that Datasette is "stringifying" the number and it's not matching? Example: https://debian-sqlelf.fly.dev/debian/elf_symbols?_sort_desc=name&_facet=exported&exported=0

There are several places where sqlite3.OperationalError is caught as part of an exception handler which catches multiple exceptions, but is then caught again immediately afterwards by a dedicated exception handler.

Because the exception will be caught by the first handler, the logic in the second handler is unreachable and will never be executed. If this is intended behavior, the second handler can be removed. If this is not intended, and the second handler should be the one that catches this exception, then sqlite3.OperationalError should be removed from the tuple of exceptions in the first handler.

This issue was found via a CodeQL query on the repository, and I've listed the occurrences found by the query below. There may be other instances of this issue in the code that were not surfaced by the query. I'd be happy to share the query if others would like to view or run it.

One example:

https://github.com/simonw/datasette/blob/452a587e236ef642cbc6ae345b58767ea8420cb5/datasette/views/database.py#L534-L537

Other instances:

https://github.com/simonw/datasette/blob/main/datasette/views/base.py#L266-L270 https://github.com/simonw/datasette/blob/main/datasette/views/base.py#L452-L456

I thought this would never happen, but now that I'm deep in the weeds of running SQLite in production for Datasette Cloud I'm starting to reconsider my policy of only supporting SQLite.

Some of the factors making me think PostgreSQL support could be worth the effort: - Serverless. I'm getting increasingly excited about writable-database use-cases for Datasette. If it could talk to PostgreSQL then users could easily deploy it on Heroku or other serverless providers that can talk to a managed RDS-style PostgreSQL. - Existing databases. Plenty of organizations have PostgreSQL databases. They can export to SQLite using db-to-sqlite but that's a pretty big barrier to getting started - being able to run datasette postgresql://connection-string and start trying it out would be a massively better experience. - Data size. I keep running into use-cases where I want to run Datasette against many GBs of data. SQLite can do this but PostgreSQL is much more optimized for large data, especially given the existence of tools like Citus. - Marketing. Convincing people to trust their data to SQLite is potentially a big barrier to adoption. Even if I've convinced myself it's trustworthy I still have to convince everyone else. - It might not be that hard? If this required a ground-up rewrite it wouldn't be worth the effort, but I have a hunch that it may not be too hard - most of the SQL in Datasette should work on both databases since it's almost all portable SELECT statements. If Datasette did DML this would be a lot harder, but it doesn't. - Plugins! This feels like a natural surface for a plugin - at which point people could add MySQL support and suchlike in the future.

The above reasons feel strong enough to justify a prototype.

When a column named 'value' is included there are no suggested facets is shown as the query uses an alias of 'value'.

https://github.com/simonw/datasette/blob/452a587e236ef642cbc6ae345b58767ea8420cb5/datasette/facets.py#L168-L174

Currently the following is shown (from https://latest.datasette.io/fixtures/facetable)

When I add a column named 'value' only the JSON facets are processed.

I think that not using aliases could be a solution (except if someone wants to use a column named count(*) though this seems to be unlikely). I'll open a PR with that.

There is also a TODO with a similar question in the same file. I have not looked into that yet.

https://github.com/simonw/datasette/blob/452a587e236ef642cbc6ae345b58767ea8420cb5/datasette/facets.py#L512

No matter what I do, I'm getting this error:

$ datasette Traceback (most recent call last): File "/Users/honza/Library/Caches/pypoetry/virtualenvs/juniorguru-Lgaxwd2n-py3.11/bin/datasette", line 5, in <module> from datasette.cli import cli File "/Users/honza/Library/Caches/pypoetry/virtualenvs/juniorguru-Lgaxwd2n-py3.11/lib/python3.11/site-packages/datasette/cli.py", line 6, in <module> from click_default_group import DefaultGroup ModuleNotFoundError: No module named 'click_default_group'

I have datasette in my dependencies like this:

toml [tool.poetry.group.dev.dependencies] datasette = {version = "1.0a7", allow-prereleases = true}

I had the latest regular version (not pre-release) there originally, but the result was the same:

toml [tool.poetry.group.dev.dependencies] datasette = "0.64.5"

Full pyproject.toml is at https://github.com/honzajavorek/junior.guru/ Previously datasette worked for me, but I guess something had to upgrade and now I can't even launch it.

https://github.com/simonw/datasette/blob/452a587e236ef642cbc6ae345b58767ea8420cb5/datasette/utils/asgi.py#L137-L139

In GET requests you can do ?foo=1&foo=2 - you can do the same in POST requests, but the dict() call here eliminates those duplicates.

You can't even try calling post_body() and implement your own custom parsing because of: - #2204

This code here:

https://github.com/simonw/datasette/blob/452a587e236ef642cbc6ae345b58767ea8420cb5/datasette/utils/asgi.py#L127-L135

It consumes the messages, which means if you try to call it a second time you won't be able to get at the body.

This is efficient - we don't end up with a request object property with potentially megabytes of content that we never look at again - but it's inconvenient for cases like middleware or functions where we don't know if the body has been consumed yet or not.

Potential solution: set request._body the first time it is called, and return that on subsequent calls.

Potential optimization: only do this for bodies that are shorter than a certain threshold - maybe 1MB - and raise an exception if you attempt to call post_body() multiple times against one of those larger bodies.

I'm a bit nervous about that option though, since it could result in errors that don't show up in testing but do show up in production.

Thanks again for such a powerful project.

For deploying to cloudrun from github actions, I'd like to create a service account with minimal permissions.

It would be great to document what those minimum permission that need to be set in the IAM.

Hi, I'm not sure if this is the right repo for this issue.

I'm using datasette with the parquet (to read a duckdb), and jellyfish plugins. Both work perfectly.

Now I need to create a simple plugin that uses the python rouge package and returns a similarity score (similarly to how the jellyfish plugin works). If I create a custom plugin, even the example hello_world one, copied directly from the tutorial, I get the following error: duckdb.duckdb.CatalogException: Catalog Error: Scalar Function with name hello_world does not exist!

Since the jellyfish plugin doesn't do anything more complex, I'm wondering if there is some other kind of issue with my setup.

It would also be interesting to try out the SQL hint mode, which can autocomplete against tables and columns. This demo shows how to configure that: https://codemirror.net/mode/sql/

Some missing documentation: https://stackoverflow.com/questions/20023381/codemirror-how-add-tables-to-sql-hint Originally posted by @simonw in https://github.com/simonw/datasette/issues/948#issuecomment-679355426

Hi, datasette version: "0.26.2" extensions: spatialite: "4.4.0-RC0" sqlite version: "3.22.0"

I have a table of airports with latitude and longitude columns. I've added spatialite (with KNN support). After creating the db using csvs-to-sqlit, I run these commands to setup the spatialite tables:

``` conn.execute('SELECT InitSpatialMetadata(1)')

conn.execute("SELECT AddGeometryColumn('airports', 'point_geom', 4326, 'POINT', 2);")

conn.execute('''UPDATE airports SET point_geom = GeomFromText('POINT('||"longitude"||' '||"latitude"||')',4326);''')

conn.execute("SELECT CreateSpatialIndex('airports', 'point_geom');") ```

I'm attempting to create a canned query and have this in my metadata.json file: "find_airports_nearest_to_point":{ "sql":"SELECT a.pos AS rank, b.id, b.name, b.country, b.latitude AS latitude, b.longitude AS longitude, a.distance / 1000.0 AS dist_km FROM KNN AS a JOIN airports AS b ON (b.rowid = a.fid) WHERE f_table_name = \"airports\" AND ref_geometry = MakePoint( :Long , :Lat ) AND max_items = 10;"} which doesn't seem to perform the templating of the name parameters correctly and I get no results.

Have also tired: MakePoint( || :Long || , || :Lat || ) which returns this error: near "||": syntax error

However I cannot seem to find the correct combination of named parameter syntax (:Lat) or sqlite concatenation operator to make it work. Any ideas if using named parameters inside functions is supported?

Thanks Darren

https://datasette.io/discord leads to https://discord.com/invite/ktd74dm5mw and returns the following:

this page

is using about 400 mb in firefox 97 on mac os x. if you download the html for the page, it's about 11mb and if you get the csv for the data its about 1mb.

it's using over a 1G on chrome 99.

i found this because, i was trying to figure out why editing the SQL was getting very slow.

I found the link to the datasette discord channel via this query.

The following video should be self explanatory:

https://github.com/simonw/datasette/assets/1892194/8cd33e88-bcaa-41f3-9818-ab4d589c3f02

Link for reference: https://discord.com/invite/ktd74dm5mw

Exception: Datasette no longer accepts plugin configuration in --metadata. Move your "plugins" configuration blocks to a separate file - we suggest calling that datasette..json - and start Datasette with datasette -c datasette..json. See https://docs.datasette.io/en/latest/configuration.html for more details.

I think we should link directly to documentation that tells people how to perform this upgrade.

Originally posted by @simonw in https://github.com/simonw/datasette/issues/2190#issuecomment-1759947021

I upgraded my dependencies, then ran into this problem running datasette inspect:

env/lib/python3.9/site-packages/datasette/cli.py", line 6, in <module> from click_default_group import DefaultGroup ModuleNotFoundError: No module named 'click_default_group'

Turns out the released version of datasette still depends on click-default-group-wheel, so click-default-group doesn't get installed/recognized:

$ virtualenv venv $ source venv/bin/activate $ pip install datasette $ pip list | grep click-default-group click-default-group 1.2.4 click-default-group-wheel 1.2.3 $ python -c "from click_default_group import DefaultGroup" Traceback (most recent call last): File "<string>", line 1, in <module> ModuleNotFoundError: No module named 'click_default_group' $ pip install --force-reinstall click-default-group ... ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts. datasette 0.64.4 requires click-default-group-wheel>=1.2.2, which is not installed. Successfully installed click-8.1.7 click-default-group-1.2.4

Using each of python -m datasette counties.db -m metadata.yml --load-extension=SpatiaLite

and

python -m datasette counties.db --load-extension="C:\Windows\System32\mod_spatialite.dll"

and

python -m datasette counties.db --load-extension=C:\Windows\System32\mod_spatialite.dll

I got the error:

``` File "C:\Users\m3n7es\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.11_qbz5n2kfra8p0\LocalCache\local-packages\Python311\site-packages\datasette\database.py", line 209, in in_thread self.ds._prepare_connection(conn, self.name) File "C:\Users\m3n7es\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.11_qbz5n2kfra8p0\LocalCache\local-packages\Python311\site-packages\datasette\app.py", line 596, in _prepare_connection conn.execute("SELECT load_extension(?, ?)", [path, entrypoint]) sqlite3.OperationalError: The specified module could not be found.

```

I finally tried modifying the code in app.py to read:

``` def _prepare_connection(self, conn, database): conn.row_factory = sqlite3.Row conn.text_factory = lambda x: str(x, "utf-8", "replace") if self.sqlite_extensions: conn.enable_load_extension(True) for extension in self.sqlite_extensions: # "extension" is either a string path to the extension # or a 2-item tuple that specifies which entrypoint to load. #if isinstance(extension, tuple): # path, entrypoint = extension # conn.execute("SELECT load_extension(?, ?)", [path, entrypoint]) #else: conn.execute("SELECT load_extension('C:\Windows\System32\mod_spatialite.dll')")

``` At which point the counties example worked.

Is there a correct way to install/use the extension on Windows? My method will cause issues if there's a second extension to be used.

On an unrelated note, my next step is to figure out how to write a query across the two loaded databases supplied from the command line: python -m datasette rm_toucans_23_10_07.db counties.db -m metadata.yml --load-extension=SpatiaLite

Original title: Perhaps metadata should all live in a _metadata in-memory database

Inspired by #1150 - metadata should be exposed as an API, and for large Datasette instances that API may need to be paginated. So why not expose it through an in-memory database table?

One catch to this: plugins. #860 aims to add a plugin hook for metadata. But if the metadata comes from an in-memory table, how do the plugins interact with it?

The need to paginate over metadata does make a plugin hook that returns metadata for an individual table seem less wise, since we don't want to have to do 10,000 plugin hook invocations to show a list of all metadata.

If those plugins write directly to the in-memory table how can their contributions survive the server restarting?

... which raises the challenge that datasette publish doesn't yet know what to do with a config file!

Originally posted by @simonw in https://github.com/simonw/datasette/issues/2194#issuecomment-1730259871

I've started to encounter a bug where queries to tables inside named in-memory databases sometimes trigger server hangs.

I'm still trying to figure out what's going on here - on one occasion I managed to Ctrl+C the server and saw an exception that mentioned a thread lock, but usually hitting Ctrl+C does nothing and I have to kill -9 the PID instead.

This is all running on my M2 Mac.

I've seen the bug in the Datasette 1.0 alphas and in Datasette 0.64.3 - but reverting to 0.61 appeared to fix it.

Got this running tests against Python 3.11.

../../../.local/share/virtualenvs/datasette-big-local-6Yn-280V/lib/python3.11/site-packages/datasette/app.py:14: in <module> import pkg_resources ../../../.local/share/virtualenvs/datasette-big-local-6Yn-280V/lib/python3.11/site-packages/pkg_resources/__init__.py:121: in <module> warnings.warn("pkg_resources is deprecated as an API", DeprecationWarning) E DeprecationWarning: pkg_resources is deprecated as an API

I ran with pytest -Werror --pdb -x to get the debugger for that warning, but it turned out searching the code worked better. It's used in these two places:

https://github.com/simonw/datasette/blob/5890a20c374fb0812d88c9b0ef26a838bfa06c76/datasette/plugins.py#L43-L50

https://github.com/simonw/datasette/blob/5890a20c374fb0812d88c9b0ef26a838bfa06c76/datasette/app.py#L1037

https://github.com/simonw/datasette/actions/runs/6266449018/job/17017460074

This is a bit of a mystery, I don't think I've changed anything recently that could have broken this.

That passed on 3.8 but should have failed: https://github.com/simonw/datasette/actions/runs/6266341481/job/17017099801 - the "Test DATASETTE_LOAD_PLUGINS" test shows errors but did not fail the CI run.

Originally posted by @simonw in https://github.com/simonw/datasette/issues/2057#issuecomment-1730201226

When running datasette serve with the --reload flag, the serve command is picked up as a file argument:

$ datasette serve --reload test_db Starting monitor for PID 13574. Error: Invalid value for '[FILES]...': Path 'serve' does not exist. Press ENTER or change a file to reload.

If a 'serve' file is created it launches properly (albeit with an empty database called serve):

$ touch serve; datasette serve --reload test_db Starting monitor for PID 13628. INFO: Started server process [13628] INFO: Waiting for application startup. INFO: Application startup complete. INFO: Uvicorn running on http://127.0.0.1:8001 (Press CTRL+C to quit)

Version (running from HEAD on main):

$ datasette --version datasette, version 1.0a2

This issue appears to have existed for awhile as https://github.com/simonw/datasette/issues/1380#issuecomment-953366110 mentions the error in a different context.

I'm happy to debug and land a patch if it's welcome.

There's a ton of tools/languages that compile to SQL, which may be nice in Datasette. Some examples:

• Logica https://logica.dev
• PRQL https://prql-lang.org
• Malloy, but not sure if it works with SQLite? https://github.com/malloydata/malloy

It would be cool if plugins could extend Datasette to use these languages, in both the code editor and API usage.

A few things I'd imagine a datasette-prql or datasette-logica plugin would do:

• prql= instead of sql=
• Code editor support (syntax highlighting, autocomplete)
• Hide/show SQL

Sometimes a plugin may want to add content to an existing default template - for example datasette-search-all adds a new search box at the top of index.html. I also want datasette-upload-csvs to add a CTA on the database.html page: https://github.com/simonw/datasette-upload-csvs/issues/18

Currently plugins can do this by providing a new version of the index.html template - but if multiple plugins try to do that only one of them will succeed.

It would be better if there were known areas of those templates which plugins could add additional content to, such that multiple plugins can use the same spot.

Hi, is there any way to use datasette for serving json only without displaying webpage? I've tried to search about this in documentation but didn't get any information

The cause of this bug: - https://github.com/simonw/datasette-export-notebook/issues/17

Is that datasette-export-notebook was consulting data["filtered_table_rows_count"] in the render output plugin function in order to show the total number of rows that would be exported.

That field is no longer available by default - the "count" field is only available if ?_extra=count was passed.

It would be useful if plugins like this could access the total count on demand, should they need to.

This made me think. That {"$env": "ENV_VAR"} hack was introduced back here:

• https://github.com/simonw/datasette/issues/538

The problem it was solving was that metadata was visible to everyone with access to the instance at /-/metadata but plugins clearly needed a way to set secret settings.

Now that this stuff is moving to config, we have some decisions to make:

1. Add /-/config to let people see the configuration of their instance, and keep the $env trick for secret settings.
2. Say all configuration aside from metadata is secret and make $env optional or ditch it entirely.
3. Allow plugins to announce which of their configuration options are secret so we can automatically redact them from /-/config

I've found /-/metadata extraordinarily useful as a user of Datasette - it really helps me understand exactly what's going on if I run into any problems with a plugin, if I can quickly check what the settings look like.

So I'm leaning towards option 1 or 3.

Originally posted by @simonw in https://github.com/simonw/datasette/pull/2183#discussion_r1325076924

Also refs: - #2093

Very often I get tripped up when trying to configure my Datasette instances. For example: if I want to change the port my app listen too, do I do that with a CLI flag, a --setting flag, inside metadata.json, or an env var? If I want to up the time limit of SQL statements, is that under metadata.json or a setting? Where does my plugin configuration go?

Normally I need to look it up in Datasette docs, and I quickly find my answer, but the number of places where "config" goes it overwhelming.

• Flat CLI flags like --port, --host, --cors, etc.
• --setting, like default_page_size, sql_time_limit_ms etc
• Inside metadata.json, including plugin configuration

Typically my Datasette deploys are extremely long shell commands, with multiple --setting and other CLI flags.

Proposal: Consolidate all "config" into datasette.toml

I propose that we add a new datasette.toml that combines "settings", "metadata", and other common CLI flags like --port and --cors into a single file. It would be similar to "Cargo.toml" in Rust projects, "package.json" in Node projects, and "pyproject.toml" in Python, etc.

A sample of what it could look like:

```toml

"top level" configuration that are currently CLI flags on datasette serve

[config] port = 8020 host = "0.0.0.0" cors = true

replaces multiple --setting flags

[settings] base_url = "/app/datasette/" default_allow_sql = true sql_time_limit_ms = 3500

replaces metadata.json.

The contents of datasette-metadata.json could be defined in this file instead, but supporting separate files is nice (since those are easy to machine-generate)

[metadata] include="./datasette-metadata.json"

plugin-specific

[plugins] [plugins.datasette-auth-github] client_id = {env = "DATASETTE_AUTH_GITHUB_CLIENT_ID"} client_secret = {env = "GITHUB_CLIENT_SECRET"}

[plugins.datasette-cluster-map]

latitude_column = "lat" longitude_column = "lon" ```

Pros

• Instead of multiple files and CLI flags, everything could be in one tidy file
• Editing config in a separate file is easier than editing CLI flags, since you don't have to kill a process + edit a command every time
• New users will know "just edit my datasette.toml instead of needing to learn metadata + settings + CLI flags
• Better dev experience for multiple environment. For example, could have datasette -c datasette-dev.toml for local dev environments (enables SQL, debug plugins, long timeouts, etc.), and a datasette -c datasette-prod.toml for "production" (lower timeouts, less plugins, monitoring plugins, etc.)

Cons

• Yet another config-management system. Now Datasette users will need to know about metadata, settings, CLI flags, and datasette.toml. However with enough documentation + announcements + examples, I think we can get ahead of it.
• If toml is chosen, would need to add a toml parser for Python version <3.11
• Multiple sources of config require priority. For example: Would --setting default_allow_sql off override the value inside [settings]? What about --port?

Other Notes

Toml

I chose toml over json because toml supports comments. I chose toml over yaml because Python 3.11 has builtin support for it. I also find toml easier to work with since it doesn't have the odd "gotchas" that YAML has ("ex 3.10 resolving to 3.1, Norway NO resolving to false, etc.). It also mimics pyproject.toml which is nice. Happy to change my mind about this however

Plugin config will be difficult

Plugin config is currently in metadata.json in two places:

1. Top level, under "plugins.[plugin-name]". This fits well into datasette.toml as [plugins.plugin-name]
2. Table level, under "databases.[db-name].tables.[table-name].plugins.[plugin-name]. This doesn't fit that well into datasette.toml, unless it's nested under [metadata]?

Extensions, static, one-off plugins?

We could also include equivalents of --plugins-dir, --static, and --load-extension into datasette.toml, but I'd imagine there's a few security concerns there to think through.

Explicitly list with plugins to use?

I believe Datasette by default will load all install plugins on startup, but maybe datasette.toml can specify a list of plugins to use? For example, a dev version of datasette.toml can specify datasette-pretty-traces, but the prod version can leave it out

There's a long justification for why this might be needed here: - https://github.com/simonw/datasette-auth-tokens/issues/10#issuecomment-1701820001

Short version: it would be neat if it was possible to stash some data on the request object such that a later plugin/middleware-type-thing could use that to influence the final returned response - similar to the kinds of things you can do with Django middleware.

The asgi_wrapper() mechanism doesn't have access to the request or response objects - it gets scope and can mess around with receive and send, but those are pretty low-level primitives.

Since Datasette has well-defined request and response objects now it might be nice to have a middleware layer that can manipulate those directly.

In building Datasette Cloud we realized that a bunch of the features we are building need a way of resolving an actor ID to the actual actor, in order to display something more interesting than just an integer ID.

Social plugins in particular need this - comments by X, CSV uploaded by X, that kind of thing.

I think the solution is a new plugin hook: actors_from_ids(datasette, ids) which can return a list of actor dictionaries.

The default implementation can return [{"id": "..."}] for the IDs passed to it.

Pluggy has a firstresult=True option which is relevant here, since this is the first plugin hook we will have implemented where only one plugin should provide an answer.

I'm going to try and get this working through an environment variable, so that you can start Datasette and it will only load a subset of plugins including those that use the register_commands() hook.

Initial research on this: - https://github.com/pytest-dev/pluggy/issues/422

Spotted this problem while working on this plugin: - https://github.com/simonw/datasette-public

It's possible to make a table public to any users - but then you may end up with situations like this:

That table is public, but the foreign key links go to tables that are NOT public.

We're also leaking the names of the values in those private tables here, which we shouldn't do. So this is a tiny bit of an information leak.

Since this only affects people who have configured a table to be public that has foreign keys to a table that is private I don't think this is worth issuing a vulnerability report about - I very much doubt anyone is running Datasette configured in a way that could result in problems because of this.

This test here: https://github.com/simonw/datasette/blob/fbcb103c0cb6668018ace539a01a6a1f156e8d6a/tests/test_api.py#L1011-L1020

It failed for me like this:

E AssertionError: assert [('normal', False), ('sqlite_stat1', True), ('sqlite_stat4', True)] in ([('normal', False), ('sqlite_stat1', True)],)

Looks like some builds of SQLite include a sqlite_stat4 table.

I unfortunately was not successful to deploy to fly.io. Please see the details above of the three scenarios that I took. I am also new to datasette.

Failed to deploy. Attaching logs: 1. Tried with an app created via flyctl apps create frosty-fog-8565 and the ran datasette publish fly covid.db --app frosty-fog-8565 ``` Deploying frosty-fog-8565 ==> Validating app configuration --> Validating app configuration done Services TCP 80/443 ⇢ 8080

Error error connecting to docker: An unknown error occured.

Traceback (most recent call last): File "c:\users\grott\anaconda3\lib\runpy.py", line 193, in _run_module_as_main "main", mod_spec) File "c:\users\grott\anaconda3\lib\runpy.py", line 85, in _run_code exec(code, run_globals) File "C:\Users\grott\Anaconda3\Scripts\datasette.exe__main__.py", line 7, in <module> File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 829, in call return self.main(args, kwargs) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 782, in main rv = self.invoke(ctx) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1066, in invoke return ctx.invoke(self.callback, ctx.params) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 610, in invoke return callback(args, **kwargs) File "c:\users\grott\anaconda3\lib\site-packages\datasette_publish_fly__init__.py", line 156, in fly "--remote-only", File "c:\users\grott\anaconda3\lib\contextlib.py", line 119, in exit next(self.gen) File "c:\users\grott\anaconda3\lib\site-packages\datasette\utils__init__.py", line 451, in temporary_docker_directory tmp.cleanup() File "c:\users\grott\anaconda3\lib\tempfile.py", line 811, in cleanup _shutil.rmtree(self.name) File "c:\users\grott\anaconda3\lib\shutil.py", line 516, in rmtree return _rmtree_unsafe(path, onerror) File "c:\users\grott\anaconda3\lib\shutil.py", line 395, in _rmtree_unsafe _rmtree_unsafe(fullname, onerror) File "c:\users\grott\anaconda3\lib\shutil.py", line 404, in _rmtree_unsafe onerror(os.rmdir, path, sys.exc_info()) File "c:\users\grott\anaconda3\lib\shutil.py", line 402, in _rmtree_unsafe os.rmdir(path) PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\Users\grott\AppData\Local\Temp\tmpgcm8cz66\frosty-fog-8565' ```

1. Tried also with an app that gets autogenerate when running flyctl launch. This also generates the .toml file. Ran then datasette publish fly covid.db --app dark-feather-168 but different error now ```Deploying dark-feather-168 ==> Validating app configuration

Error not possible to validate configuration: server returned Post "https://api.fly.io/graphql": unexpected EOF

Traceback (most recent call last): File "c:\users\grott\anaconda3\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec) exec(code, run_globals) File "C:\Users\grott\Anaconda3\Scripts\datasette.exe__main__.py", line 7, in <module> File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 829, in call return self.main(args, kwargs) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 782, in main rv = self.invoke(ctx) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1066, in invoke return ctx.invoke(self.callback, ctx.params) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 610, in invoke return callback(args, **kwargs) File "c:\users\grott\anaconda3\lib\site-packages\datasette_publish_fly__init__.py", line 156, in fly "--remote-only", File "c:\users\grott\anaconda3\lib\contextlib.py", line 119, in exit next(self.gen) File "c:\users\grott\anaconda3\lib\site-packages\datasette\utils__init__.py", line 451, in temporary_docker_directory tmp.cleanup() File "c:\users\grott\anaconda3\lib\tempfile.py", line 811, in cleanup _shutil.rmtree(self.name) File "c:\users\grott\anaconda3\lib\shutil.py", line 516, in rmtree return _rmtree_unsafe(path, onerror) File "c:\users\grott\anaconda3\lib\shutil.py", line 395, in _rmtree_unsafe _rmtree_unsafe(fullname, onerror) File "c:\users\grott\anaconda3\lib\shutil.py", line 404, in _rmtree_unsafe onerror(os.rmdir, path, sys.exc_info()) File "c:\users\grott\anaconda3\lib\shutil.py", line 402, in _rmtree_unsafe os.rmdir(path) PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\Users\grott\AppData\Local\Temp\tmpnoyewcre\dark-feather-168' ```

These are also the contents of the generated .toml file in 2 scenario:

```

fly.toml file generated for dark-feather-168 on 2021-09-28T20:35:44+02:00

app = "dark-feather-168"

kill_signal = "SIGINT" kill_timeout = 5 processes = []

[env]

[experimental] allowed_public_ports = [] auto_rollback = true

[[services]] http_checks = [] internal_port = 8080 processes = ["app"] protocol = "tcp" script_checks = []

[services.concurrency] hard_limit = 25 soft_limit = 20 type = "connections"

[[services.ports]] handlers = ["http"] port = 80

[[services.ports]] handlers = ["tls", "http"] port = 443

[[services.tcp_checks]] grace_period = "1s" interval = "15s" restart_limit = 6 timeout = "2s" ```

1. But also trying datasette package covid.db to create a local DOCKERFILE to later try to push it via flyctl deploy fails as well.

```[+] Building 147.3s (11/11) FINISHED => [internal] load build definition from Dockerfile 0.2s => => transferring dockerfile: 396B 0.0s => [internal] load .dockerignore 0.1s => => transferring context: 2B 0.0s => [internal] load metadata for docker.io/library/python:3.8 4.7s => [auth] library/python:pull token for registry-1.docker.io 0.0s => [internal] load build context 0.1s => => transferring context: 82.37kB 0.0s => [1/5] FROM docker.io/library/python:3.8@sha256:530de807b46a11734e2587a784573c12c5034f2f14025f838589e6c0e3 108.3s => => resolve docker.io/library/python:3.8@sha256:530de807b46a11734e2587a784573c12c5034f2f14025f838589e6c0e3b5 0.0s => => sha256:56182bcdf4d4283aa1f46944b4ef7ac881e28b4d5526720a4e9ba03a4730846a 2.22kB / 2.22kB 0.0s => => sha256:955615a668ce169f8a1443fc6b6e6215f43fe0babfb4790712a2d3171f34d366 54.93MB / 54.93MB 21.6s => => sha256:911ea9f2bd51e53a455297e0631e18a72a86d7e2c8e1807176e80f991bde5d64 10.87MB / 10.87MB 15.5s => => sha256:530de807b46a11734e2587a784573c12c5034f2f14025f838589e6c0e3b5c5b6 1.86kB / 1.86kB 0.0s => => sha256:ff08f08727e50193dcf499afc30594c47e70cc96f6fcfd1a01240524624264d0 8.65kB / 8.65kB 0.0s => => sha256:2756ef5f69a5190f4308619e0f446d95f5515eef4a814dbad0bcebbbbc7b25a8 5.15MB / 5.15MB 6.4s => => sha256:27b0a22ee906271a6ce9ddd1754fdd7d3b59078e0b57b6cc054c7ed7ac301587 54.57MB / 54.57MB 37.7s => => sha256:8584d51a9262f9a3a436dea09ba40fa50f85802018f9bd299eee1bf538481077 196.45MB / 196.45MB 82.3s => => sha256:524774b7d3638702fe9ae0ea3fcfb81b027dfd75cc2fc14f0119e764b9543d58 6.29MB / 6.29MB 26.6s => => extracting sha256:955615a668ce169f8a1443fc6b6e6215f43fe0babfb4790712a2d3171f34d366 5.4s => => sha256:9460f6b75036e38367e2f27bb15e85777c5d6cd52ad168741c9566186415aa26 16.81MB / 16.81MB 40.5s => => extracting sha256:2756ef5f69a5190f4308619e0f446d95f5515eef4a814dbad0bcebbbbc7b25a8 0.6s => => extracting sha256:911ea9f2bd51e53a455297e0631e18a72a86d7e2c8e1807176e80f991bde5d64 0.6s => => sha256:9bc548096c181514aa1253966a330134d939496027f92f57ab376cd236eb280b 232B / 232B 40.1s => => extracting sha256:27b0a22ee906271a6ce9ddd1754fdd7d3b59078e0b57b6cc054c7ed7ac301587 5.8s => => sha256:1d87379b86b89fd3b8bb1621128f00c8f962756e6aaaed264ec38db733273543 2.35MB / 2.35MB 41.8s => => extracting sha256:8584d51a9262f9a3a436dea09ba40fa50f85802018f9bd299eee1bf538481077 18.8s => => extracting sha256:524774b7d3638702fe9ae0ea3fcfb81b027dfd75cc2fc14f0119e764b9543d58 1.2s => => extracting sha256:9460f6b75036e38367e2f27bb15e85777c5d6cd52ad168741c9566186415aa26 2.9s => => extracting sha256:9bc548096c181514aa1253966a330134d939496027f92f57ab376cd236eb280b 0.0s => => extracting sha256:1d87379b86b89fd3b8bb1621128f00c8f962756e6aaaed264ec38db733273543 0.8s => [2/5] COPY . /app 2.3s => [3/5] WORKDIR /app 0.2s => [4/5] RUN pip install -U datasette 26.9s => [5/5] RUN datasette inspect covid.db --inspect-file inspect-data.json 3.1s => exporting to image 1.2s => => exporting layers 1.2s => => writing image sha256:b5db0c205cd3454c21fbb00ecf6043f261540bcf91c2dfc36d418f1a23a75d7a 0.0s

Use 'docker scan' to run Snyk tests against images to find vulnerabilities and learn how to fix them Traceback (most recent call last): "main", mod_spec) File "c:\users\grott\anaconda3\lib\runpy.py", line 85, in _run_code exec(code, run_globals) File "C:\Users\grott\Anaconda3\Scripts\datasette.exe__main__.py", line 7, in <module> File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 829, in call return self.main(args, kwargs) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 782, in main rv = self.invoke(ctx) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1259, in invoke return _process_result(sub_ctx.command.invoke(sub_ctx)) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 1066, in invoke return ctx.invoke(self.callback, ctx.params) File "c:\users\grott\anaconda3\lib\site-packages\click\core.py", line 610, in invoke return callback(args, **kwargs) File "c:\users\grott\anaconda3\lib\site-packages\datasette\cli.py", line 283, in package call(args) File "c:\users\grott\anaconda3\lib\contextlib.py", line 119, in exit next(self.gen) File "c:\users\grott\anaconda3\lib\site-packages\datasette\utils__init__.py", line 451, in temporary_docker_directory tmp.cleanup() File "c:\users\grott\anaconda3\lib\tempfile.py", line 811, in cleanup _shutil.rmtree(self.name) File "c:\users\grott\anaconda3\lib\shutil.py", line 516, in rmtree return _rmtree_unsafe(path, onerror) File "c:\users\grott\anaconda3\lib\shutil.py", line 395, in _rmtree_unsafe _rmtree_unsafe(fullname, onerror) File "c:\users\grott\anaconda3\lib\shutil.py", line 404, in _rmtree_unsafe onerror(os.rmdir, path, sys.exc_info()) File "c:\users\grott\anaconda3\lib\shutil.py", line 402, in _rmtree_unsafe os.rmdir(path) PermissionError: [WinError 32] The process cannot access the file because it is being used by another process: 'C:\Users\grott\AppData\Local\Temp\tmpkb27qid3\datasette'```

This came up in discussion over: - https://github.com/simonw/datasette/pull/2174

I've used this pattern twice now: https://til.simonwillison.net/datasette/register-new-plugin-hooks - in datasette-graphql and datasette-low-disk-space-hook. I should describe the pattern on https://docs.datasette.io/en/stable/writing_plugins.html

I noticed that a token with execute-sql permission alone did not work, because it was not allowed to view the instance of the database.

The return type isn't documented here: https://github.com/simonw/datasette/blob/4c3ef033110407f3b3dbce501659d523724985e0/docs/internals.rst#L327-L350

On inspecting the code I'm not 100% sure if it's possible for this. method to return None, or if it can only return True or False. Need to confirm that.

https://github.com/simonw/datasette/blob/4c3ef033110407f3b3dbce501659d523724985e0/datasette/app.py#L822C15-L853

The current _internal database is used by Datasette core to cache info about databases/tables/columns/foreign keys of databases in a Datasette instance. It's a temporary database created at startup, that can only be seen by the root user. See an example _internal DB here, after logging in as root.

The current _internal database has a few rough edges:

• It's part of datasette.databases, so many plugins have to specifically exclude _internal from their queries examples here
• It's only used by Datasette core and can't be used by plugins or 3rd parties
• It's created from scratch at startup and stored in memory. Why is fine, the performance is great, but persistent storage would be nice.

Additionally, it would be really nice if plugins could use this _internal database to store their own configuration, secrets, and settings. For example:

• datasette-auth-tokens creates a _datasette_auth_tokens table to store auth token metadata. This could be moved into the _internal database to avoid writing to the gues database
• datasette-socrata creates a socrata_imports table, which also can be in _internal
• datasette-upload-csvs creates a _csv_progress_ table, which can be in _internal
• datasette-write-ui wants to have the ability for users to toggle whether a table appears editable, which can be either in datasette.yaml or on-the-fly by storing config in _internal

In general, these are specific features that Datasette plugins would have access to if there was a central internal database they could read/write to:

• Dynamic configuration. Changing the datasette.yaml file works, but can be tedious to restart the server every time. Plugins can define their own configuration table in _internal, and could read/write to it to store configuration based on user actions (cell menu click, API access, etc.)
• Caching. If a plugin or Datasette Core needs to cache some expensive computation, they can store it inside _internal (possibly as a temporary table) instead of managing their own caching solution.
• Audit logs. If a plugin performs some sensitive operations, they can log usage info to _internal for others to audit later.
• Long running process status. Many plugins (datasette-upload-csvs, datasette-litestream, datasette-socrata) perform tasks that run for a really long time, and want to give continue status updates to the user. They can store this info inside_internal
• Safer authentication. Passwords and authentication plugins usually store credentials/hashed secrets in configuration files or environment variables, which can be difficult to handle. Now, they can store them in _internal

Proposal

• We remove _internal from datasette.databases property.
• We add new datasette.get_internal_db() method that returns the _internal database, for plugins to use
• We add a new --internal internal.db flag. If provided, then the _internal DB will be sourced from that file, and further updates will be persisted to that file (instead of an in-memory database)
• When creating internal.db, create a new _datasette_internal table to mark it a an "datasette internal database"
• In datasette serve, we check for the existence of the _datasette_internal table. If it exists, we assume the user provided that file in error and raise an error. This is to limit the chance that someone accidentally publishes their internal database to the internet. We could optionally add a --unsafe-allow-internal flag (or database plugin) that allows someone to do this if they really want to.

New features unlocked with this

These features don't really need a standardized _internal table per-say (plugins could currently configure their own long-time storage features if they really wanted to), but it would make it much simpler to create these kinds of features with a persistent application database.

• datasette-comments : A plugin for commenting on rows or specific values in a database. Comment contents + threads + email notification info can be stored in _internal
• Bookmarks: "Bookmarking" an SQL query could be stored in _internal, or a URL link shortener
• Webhooks: If a plugin wants to either consume a webhook or create a new one, they can store hashed credentials/API endpoints in _internal

The BaseView class has a method for cascading permission checks, but it's not easily accessible to plugins.

https://github.com/simonw/datasette/blob/5eb8e9bf250b26e30b017d39a392c33973997656/datasette/views/base.py#L75-L99

This leaves plugins like datasette-graphql having to implement their own versions of this logic, which is bad: https://github.com/simonw/datasette-graphql/issues/65

First check view-database - if that says False then disallow access, if it says True then allow access. If it says None check view-instance.

This should become a supported API that plugins are encouraged to use.

I just had to write a one-off custom plugin to redirect niche-musems.com to www.niche-museums.com (https://github.com/simonw/museums/issues/21) - it would be great if this kind of thing could be handled by a configurable plugin.

https://github.com/simonw/museums/blob/6b1faf00c463b2228860d4d62d104b11935e01b1/plugins/redirect_www.py

If I have a self-hosted instance of Datasette up and running, I'd like to be able to the use the CLI to publish databases to that instance, not only Google or Heroku. Ideally there'd be a url parameter or something similar to which one could point the publish command to their instance.

Discussed with Alex this morning. We think the American spelling is fine here (it's shorter than catalogue) and that it's a slightly less lazy name than core_.

Follows: - https://github.com/simonw/datasette/issues/2157

Spotted a problem while working on this: if you grant a token access to view table for a specific table but don't also grant view database and view instance permissions, that token is useless.

This was a deliberate design decision in Datasette - it's documented on https://docs.datasette.io/en/1.0a2/authentication.html#access-permissions-in-metadata

If a user cannot access a specific database, they will not be able to access tables, views or queries within that database. If a user cannot access the instance they will not be able to access any of the databases, tables, views or queries.

I'm now second-guessing if this was a good decision.

Originally posted by @simonw in https://github.com/simonw/datasette-auth-tokens/issues/7#issuecomment-1636031702

I experimented with a prototype of this here: - https://github.com/simonw/datasette/issues/2102#issuecomment-1691037971_

Which lets me run requests as if they belonged to a specific actor like this:

bash datasette fixtures.db --get '/fixtures/facetable.json' --actor '{ "_r": { "r": { "fixtures": { "facetable": [ "vt" ] } } }, "a": "user" }'

Really useful for testing actors an _r options. Is this worth adding as a feature?

I've been thinking about what it might look like to allow command-line arguments to be used to define any of the configuration options in datasette.yml, as alternative and more convenient syntax.

Here's what I've come up with: datasette \ -s settings.sql_time_limit_ms 1000 \ -s plugins.datasette-auth-tokens.manage_tokens true \ -s plugins.datasette-auth-tokens.manage_tokens_database tokens \ mydatabase.db tokens.db Which would be equivalent to datasette.yml containing this: yaml plugins: datasette-auth-tokens: manage_tokens: true manage_tokens_database: tokens settings: sql_time_limit_ms: 1000 More details in https://github.com/simonw/datasette/issues/2143#issuecomment-1690792514

In thinking about the best way to implement https://github.com/simonw/datasette-auth-passwords/issues/6 (SQL-backed user accounts for datasette-auth-passwords) I realized that there are a few different use-cases where a plugin might want to store data that isn't visible to regular Datasette users:

• Storing password hashes
• Storing API tokens
• Storing secrets that are used for data import integrations (secrets for talking to the Twitter API for example)

Idea: allow one or more private database files to be attached to Datasette, something like this:

datasette github.db linkedin.db -s secrets.db -m metadata.yml

The secrets.db file would not be visible using any of the Datasette's usual interface or API routes - but plugins would be able to run queries against it.

So datasette-auth-passwords might then be configured like this:

yaml plugins: datasette-auth-passwords: database: secrets sql: "select password_hash from passwords where username = :username" The plugin could even refuse to operate against a database that hadn't been loaded as a secret database.

I use datasette in an iframe inside another HTML file that contains other ways to represent my data (mostly leaflets maps built with R on summarized data), and the datasette iframe is a tab in that page.

This particular use prevents users to access the full URLs of their datasette views and queries, which is a shame because the way datasette handles URLs to make every view or query easy to share is awesome. I know how to get the URL from the context menu of my browser, but I don't think many visitors would do it or even notice that datasette uses permalinks for pretty much every action they do. Would it be possible to add a "Share link" button to the interface, either in datasette itself or in a plugin?

Metadata in Datasette is a really powerful feature, but is a bit difficult to work with. It was initially a way to add "metadata" about your "data" in Datasette instances, like descriptions for databases/tables/columns, titles, source URLs, licenses, etc. But it later became the go-to spot for other Datasette features that have nothing to do with metadata, like permissions/plugins/canned queries.

Specifically, I've found the following problems when working with Datasette metadata:

1. Metadata cannot be updated without re-starting the entire Datasette instance.
2. The metadata.json/metadata.yaml has become a kitchen sink of unrelated (imo) features like plugin config, authentication config, canned queries
3. The Python APIs for defining extra metadata are a bit awkward (the datasette.metadata() class, get_metadata() hook, etc.)

Possible solutions

Here's a few ideas of Datasette core changes we can make to address these problems.

Re-vamp the Datasette Python metadata APIs

The Datasette object has a single datasette.metadata() method that's a bit difficult to work with. There's also no Python API for inserted new metadata, so plugins have to rely on the get_metadata() hook.

The get_metadata() hook can also be improved - it doesn't work with async functions yet, so you're quite limited to what you can do.

(I'm a bit fuzzy on what to actually do here, but I imagine it'll be very small breaking changes to a few Python methods)

Add an optional datasette_metadata table

Datasette should detect and use metadata stored in a new special table called datasette_metadata. This would be a regular table that a user can edit on their own, and would serve as a "live updating" source of metadata, than can be changed while the Datasette instance is running.

Not too sure what the schema would look like, but I'd imagine:

sql CREATE TABLE datasette_metadata( level text, target any, key text, value any, primary key (level, target) )

Every row in this table would map to a single metadata "entry".

• level would be one of "datasette", "database", "table", "column", which is the "level" the entry describes. For example, level="table" means it is metadata about a specific table, level="database" for a specific database, or level="datasette" for the entire Datasette instance.
• target would "point" to the specific object the entry metadata is about, and would depend on what level is specific.
• level="database": target would be the string name of the database that the metadata entry is about. ex "fixtures"
• level="table": target would be a JSON array of two strings. The first element would be the database name, and the second would be the table name. ex ["fixtures", "students"]
• level="column": target would be a JSON array of 3 strings: The database name, table name, and column name. Ex ["fixtures", "students", "student_id"]
• key would be the type of metadata entry the row has, similar to the current "keys" that exist in metadata.json. Ex "about_url", "source", "description", etc
• value would be the text value of be metadata entry. The literal text value of a description, about_url, column_label, etc

A quick sample:

level | target | key | value -- | -- | -- | -- datasette | NULL | title | my datasette title... db | fixtures | source | <description of my database source> table | ["fixtures", "students"] | label_column | student_name column | ["fixtures", "students", "birthdate"] | description | <description of the fixtures.students.birthdate column>

This datasette_metadata would be configured with other tools, and hopefully not manually by end users. Datasette Core could also offer a UI for editing entries in datasette_metadata, to update descriptions/columns on the fly.

Re-vamp metadata.json and move non-metadata config to another place

The motivation behind this is that it's awkward that metadata.json contains config about things that are not strictly metadata, including:

• Plugin configuration
• Authentication/permissions (ex the allow key on datasettes/databases/tables
• Canned queries. might be controversial, but in my mind, canned queries are application-specific code and configuration, and don't describe the data that exists in SQLite databases.

I think we should move these outside of metadata.json and into a different file. The datasette.json idea in #2093 may be a good solution here: plugin/permissions/canned queries can be defined in datasette.json, while metadata.json/datasette_metadata will strictly be about documenting databases/tables/columns.

I'm interested in making a plugin that saves every query that gets run to a table in the database. (I know about datasette-query-history but thought it would be good to have a server-side option.)

As far as I can tell reading the docs, there isn't really a hook setup to allow this.

Maybe I could hack it with some of the hooks that are passed requests, but that doesn't seem good.

I'm a little surprised this isn't possible, so I thought I would open an issue and see if that's a deeply considered decision or just "haven't needed it yet." I'm potentially interested in implementing the hook if the latter.

See: - https://github.com/simonw/datasette-configure-fts/issues/14 - https://github.com/simonw/datasette-auth-tokens/issues/12

It's EOL now, as of 2023-06-27 (two days ago): https://devguide.python.org/versions/

I'm going to drop Python 3.7.

Originally posted by @simonw in https://github.com/simonw/datasette/issues/1153#issuecomment-1627455892

It's not supported any more: https://devguide.python.org/versions/

I want to use isort to order imports. A few steps here:

• [x] Add a .isort.cfg file (see below)
• [x] Use isort -rc to reformat existing code
• [ ] Commit this change
• [x] Add a unit test that ensures future changes remain isort compatible

It is increasingly being useful configuration options, when it started out as purely metadata.

Could cause confusion with the --config mechanism though - maybe that should be called "settings" instead?

Split off from #274

Metadata should an optional "config" section at root, table or database level.

The TableView and RowView and DatabaseView and BaseView classes could all have a .config("key") method which knows how to resolve the hierarchy of configs.

This will allow individual tables (or databases) to set their own config settings for things like sql_time_limit_ms

Stumbled across this while experimenting with datasette-write-ui. The error I got was a 500 on the /db page:

'NoneType' object has no attribute 'encode'

Tracked it down to this code, which assembles the URL for a row page:

https://github.com/simonw/datasette/blob/943df09dcca93c3b9861b8c96277a01320db8662/datasette/utils/init.py#L120-L134

That's because tilde_encode can't handle None: https://github.com/simonw/datasette/blob/943df09dcca93c3b9861b8c96277a01320db8662/datasette/utils/init.py#L1175-L1178

For example this: https://github.com/simonw/datasette/blob/4535568f2ce907af646304d0ebce2500ebd55677/docs/authentication.rst?plain=1#L33-L35

The problem with that $ prefix is that it prevents users from copying and pasting the raw command.

https://docs.datasette.io/en/stable/authentication.html#using-the-root-actor

https://docs.datasette.io/en/latest/authentication.html#other-permissions-in-metadata says the following:

For all other permissions, you can use one or more "permissions" blocks in your metadata.

To grant access to the permissions debug tool to all signed in users you can grant permissions-debug to any actor with an id matching the wildcard * by adding this a the root of your metadata:

yaml permissions: debug-menu: id: '*'

I tried this.

My metadata.yml file looks like:

yaml permissions: debug-menu: id: '*' permissions-debug: id: '*' plugins: datasette-auth-passwords: myuser_password_hash: $env: "PASSWORD_HASH_MYUSER"

And then I run

zsh datasette -m metadata.yml tiddlywiki.db --root

And I open a session for the "root" user of datasette with the link given.

I open a private browser session and log in as "myuser" from http://127.0.0.1:8001/-/login

Then I check http://127.0.0.1:8001/-/actor which confirms that I am logged in as the "myuser" actor

json { "actor": { "id": "myuser" } }

In the session where I am logged in as "myuser" I then try to go to http://127.0.0.1:8001/-/permissions

But all I get there as the logged in user "myuser" is

Forbidden

Permission denied

And then if I check the http://127.0.0.1:8001/-/permissions as the datasette "root" user from another browser session, I see:

permissions-debug checked at 2023-08-06T16:22:58.997841 ✗ (used default)

Actor: {"id": "myuser"}

It seems that in spite of having tried to give the permissions-debug permission to the "myuser" user in my metadata.yml file, datasette does not agree that "myuser" has permission permissions-debug..

What do I need to do differently so that my "myuser" user is able to access http://127.0.0.1:8001/-/permissions ?

I think this has been a bug for a long time.

https://latest.datasette.io/ currently shows:

Those colors are based on a hash of the database name. But when you click through to https://latest.datasette.io/fixtures

It's red on all sub-pages too.

This would be really useful as it would allow us to tie in with colour schemes.

Spotted this on https://latest.datasette.io/extra_database

```html

This idea first came up here: - https://github.com/simonw/datasette/issues/2112#issuecomment-1652751140

If datasette.render_template(...) takes an optional Context subclass as an alternative to a context dictionary, I could then use dataclasses to define the context made available to specific templates - which then gives me something I can use to help document what they are.

Also refs: - https://github.com/simonw/datasette/issues/1510

This would be useful while working on: - #2114

Or... how about if the on_success_message option could define a SQL query to be executed to generate that message? Maybe on_success_message_sql.

• https://github.com/simonw/datasette/issues/2134

Hi, simon ❤️

datasette plugins --all > plugins.json could generate all plugins info. On another machine, it would be great to install all plugins just by datasette install plugins.json

The default JSON just isn't right. I find myself using ?_shape=array for almost everything I build against the API.

118 commits! https://github.com/simonw/datasette/compare/1.0a2...26be9f0445b753fb84c802c356b0791a72269f25

I just noticed that https://latest.datasette.io/fixtures/roadside_attraction_characteristics.json?_labels=on&_size=1 returns: json { "ok": true, "next": "1", "rows": [ { "rowid": 1, "attraction_id": { "value": 1, "label": "The Mystery Spot" }, "characteristic_id": { "value": 2, "label": "Paranormal" } } ], "truncated": false } But https://latest.datasette.io/fixtures.json?sql=select+rowid%2C+attraction_id%2C+characteristic_id+from+roadside_attraction_characteristics+order+by+rowid+limit+1 returns: json { "rows": [ { "rowid": 1, "attraction_id": 1, "characteristic_id": 2 } ], "columns": [ "rowid", "attraction_id", "characteristic_id" ], "ok": true, "truncated": false } The columns key in the query response is inconsistent with the table response.

This URL here:

https://latest.datasette.io/fixtures/simple_primary_key/1.json?_extras=foreign_key_tables

json { "database": "fixtures", "table": "simple_primary_key", "rows": [ { "id": "1", "content": "hello" } ], "columns": [ "id", "content" ], "primary_keys": [ "id" ], "primary_key_values": [ "1" ], "units": {}, "foreign_key_tables": [ { "other_table": "foreign_key_references", "column": "id", "other_column": "foreign_key_with_blank_label", "count": 0, "link": "/fixtures/foreign_key_references?foreign_key_with_blank_label=1" }, { "other_table": "foreign_key_references", "column": "id", "other_column": "foreign_key_with_label", "count": 1, "link": "/fixtures/foreign_key_references?foreign_key_with_label=1" }, { "other_table": "complex_foreign_keys", "column": "id", "other_column": "f3", "count": 1, "link": "/fixtures/complex_foreign_keys?f3=1" }, { "other_table": "complex_foreign_keys", "column": "id", "other_column": "f2", "count": 0, "link": "/fixtures/complex_foreign_keys?f2=1" }, { "other_table": "complex_foreign_keys", "column": "id", "other_column": "f1", "count": 1, "link": "/fixtures/complex_foreign_keys?f1=1" } ], "query_ms": 4.226590999678592, "source": "tests/fixtures.py", "source_url": "https://github.com/simonw/datasette/blob/main/tests/fixtures.py", "license": "Apache License 2.0", "license_url": "https://github.com/simonw/datasette/blob/main/LICENSE", "ok": true, "truncated": false }

That ?_extras= should be ?_extra= - plus the row JSON should be redesigned to fit the new default JSON representation.

• 2109

From Discord: https://discordapp.com/channels/823971286308356157/996877076982415491/1068227071156965486

I ended up duplicating it here: https://github.com/simonw/datasette/blob/7532feb424b1dce614351e21b2265c04f9669fe2/datasette/views/database.py#L164-L167

I think it should move to datasette.render_template() - and maybe have a renamed template variable too.

• 2109

Caused by: - #2112

https://latest.datasette.io/fixtures?sql=select+pk1%2C+pk2%2C+pk3%2C+content+from+compound_three_primary_keys+where+%22pk1%22+%3D+%3Ap0+order+by+pk1%2C+pk2%2C+pk3+limit+101&p0=b

The :p0 form field is missing. Submitting the form results in this error:

https://latest.datasette.io/_memory.csv?sql=select+blah is a blank page right now:

bash curl -I 'https://latest.datasette.io/_memory.csv?sql=select+blah' HTTP/2 200 access-control-allow-origin: * access-control-allow-headers: Authorization, Content-Type access-control-expose-headers: Link access-control-allow-methods: GET, POST, HEAD, OPTIONS access-control-max-age: 3600 content-type: text/plain; charset=utf-8 x-databases: _memory, _internal, fixtures, fixtures2, extra_database, ephemeral date: Mon, 07 Aug 2023 23:12:15 GMT server: Google Frontend

Originally posted by @simonw in https://github.com/simonw/datasette/issues/2118#issuecomment-1668688947

In the old code:

https://github.com/simonw/datasette/blob/08181823990a71ffa5a1b57b37259198eaa43e06/datasette/views/database.py#L34-L35

This name class attribute was later used by some of the plugin hooks, passed as view_name: https://github.com/simonw/datasette/blob/18dd88ee4d78fe9d760e9da96028ae06d938a85c/datasette/hookspecs.py#L50-L54

Figure out how that should work once I've refactored those classes to view functions instead.

Refs: - #2109

• 2109

This will help make the hook as robust as possible. - #2109

This will be the base that the remaining work builds on top of. Refs: - #2109

While working on: - https://github.com/simonw/datasette/pull/2118

It became clear that the render callback function documented here: https://docs.datasette.io/en/0.64.3/plugin_hooks.html#register-output-renderer-datasette

Needs to grow the ability to be told if an error occurred (an error string) and if the results were truncated (a truncated boolean).

Hi,

currently parameters in canned queries are just text fields. It would be cool to have one of the options below. Would you accept a PR doing something in this direction? (Possibly this could even work as a plugin.)

• adding facets, which would work like facets on tables or views, giving a list of selectable options (and leaving parameters as is)
• making it possible to provide a query which returns selectable values for a parameter, e.g. calendar_entries_current_instrument: sql: | select * from calendar_entries where DTEND_UNIX > UNIXEPOCH() and DTSTART_UNIX < UNIXEPOCH() + :days *24*60*60 and current = 1 and MACHINE = :instrument order by DTSTART_UNIX params: days: sql: "SELECT VALUE FROM generate_series(1, 30, 1)" # this obviously requires the corresponding sqlite extension instrument: sql: "SELECT DISTINCT MACHINE FROM calendar_entries"
• making it possible to provide a fixed list of parameters calendar_entries_current_instrument: sql: | select * from calendar_entries where DTEND_UNIX > UNIXEPOCH() and DTSTART_UNIX < UNIXEPOCH() + :days *24*60*60 and current = 1 and MACHINE = :instrument order by DTSTART_UNIX params: days: values: [1, 2, 3, 5, 10, 20, 30] instrument: values: [supermachine, crappymachine, boringmachine]

Currently looks like this: https://github.com/simonw/datasette/blob/08181823990a71ffa5a1b57b37259198eaa43e06/datasette/init.py#L1-L6

Adding __all__ = ["Permission", "Forbidden"...] would let me get rid of those # noqa comments.

I've been stuck on this for too long. I'm breaking it down into a full milestone:

https://github.com/simonw/datasette/milestone/29

Refs: - #2109

The idea here is that hitting /content without a ?sql= will show an empty result set AND default to including a bunch of extras about the list of tables in the database.

Then I won't have to think about /content and /content?sql= as separate pages any more.

A minor refactor, but it is a good starting point for this new branch. Refs: - #2109

As seen in LLM and now in sqlite-utils too: - https://github.com/simonw/sqlite-utils/issues/570

Useful for developing plugins, see tutorial at https://llm.datasette.io/en/stable/plugins/tutorial-model-plugin.html

here's a CTE query that should always fail with a syntax error:

sql with foo as (nonsense) select * from foo;

when we make this query against the default endpoint, we do indeed get a 400 status code the problem is returned to the user: https://global-power-plants.datasettes.com/global-power-plants?sql=with+foo+as+%28nonsense%29+select+*+from+foo%3B

but, if we use the csv endpoint, we get a 200 status code and no indication of a problem: https://global-power-plants.datasettes.com/global-power-plants.csv?sql=with+foo+as+%28nonsense%29+select+*+from+foo%3B

same with this bad sql

sql select a, from foo;

https://global-power-plants.datasettes.com/global-power-plants?sql=select%0D%0A++a%2C%0D%0Afrom%0D%0A++foo%3B

vs

https://global-power-plants.datasettes.com/global-power-plants.csv?sql=select%0D%0A++a%2C%0D%0Afrom%0D%0A++foo%3B

but, datasette catches this bad sql at both endpoints:

sql slect a from foo;

https://global-power-plants.datasettes.com/global-power-plants?sql=slect%0D%0A++a%0D%0Afrom%0D%0A++foo%3B https://global-power-plants.datasettes.com/global-power-plants.csv?sql=slect%0D%0A++a%0D%0Afrom%0D%0A++foo%3B

I use this nginx config: ``` location /datasette-llm { return 302 /datasette-llm/; }

location /datasette-llm/ {
  proxy_set_header Upgrade           $http_upgrade;
  proxy_set_header Connection        "Upgrade";
  proxy_http_version 1.1;
  proxy_set_header X-Real-IP         $remote_addr;
  proxy_set_header X-Forwarded-For   $proxy_add_x_forwarded_for;
  proxy_set_header X-Forwarded-Proto https;
  proxy_set_header X-Forwarded-Host  $http_host;
  proxy_set_header Host              $host;
  proxy_max_temp_file_size           0;
  proxy_pass                         http://127.0.0.1:8001/datasette-llm/;
  proxy_redirect                     http:// https://;
  proxy_buffering off;
  proxy_request_buffering off;
  proxy_set_header Origin            '';
  client_max_body_size 0;
  auth_basic                         "datasette-llm";
  auth_basic_user_file               /etc/nginx/custom-userdb;
}

Then I start datasette with this command: datasette serve --setting base_url /datasette-llm/ $(llm logs path) ``` Everything else works right, except the links in "This data as json, CSV". They get an extra URL element "datasette-llm" like this:

https://192.168.1.3:5432/datasette-llm/datasette-llm/logs.json?sql=select+*+from+_llm_migrations

https://192.168.1.3:5432/datasette-llm/datasette-llm/logs.csv?sql=select+*+from+_llm_migrations&_size=max

When I remove that extra "datasette-llm" from the URL, those links work too.

Plugins can then take advantage of this pattern, for example: - https://github.com/simonw/datasette-auth-tokens/pull/8

Maybe put it on the <tr> but probably better to go on the td.type-pk.

https://discord.com/channels/823971286308356157/823971286941302908/1120705940728066080

May I add a request to the whole metadata / settings ? Allow to pass --settings path/to/settings.json instead of having to rely exclusively on directory mode to centralize settings (this would reflect the behavior of providing metadata)

Requested here: https://discord.com/channels/823971286308356157/823971286941302908/1129034187073134642

The former datasette-insert plugin had an option ?alter=1 to auto-add new columns. Does the JSON write API also have this?

YAML configuration is much better for multi-line strings, and I'm increasingly adding configuration options to Datasette that benefit from that - fragments of HTML in description_html or SQL queries used to configure things like https://github.com/simonw/datasette-atom for example.

Rather than confusing things by showing both in the documentation, I should switch all of the default examples to use YAML instead.

Using the CSS media query prefers-color-scheme we can provide a dark-mode version of Datasette