## Changes
This is a fundamental change to how we load and process bundle
configuration. We now depend on the configuration being represented as a
`dyn.Value`. This representation is functionally equivalent to Go's
`any` (it is variadic) and allows us to capture metadata associated with
a value, such as where it was defined (e.g. file, line, and column). It
also allows us to represent Go's zero values properly (e.g. empty
string, integer equal to 0, or boolean false).
Using this representation allows us to let the configuration model
deviate from the typed structure we have been relying on so far
(`config.Root`). We need to deviate from these types when using
variables for fields that are not a string themselves. For example,
using `${var.num_workers}` for an integer `workers` field was impossible
until now (though not implemented in this change).
The loader for a `dyn.Value` includes functionality to capture any and
all type mismatches between the user-defined configuration and the
expected types. These mismatches can be surfaced as validation errors in
future PRs.
Given that many mutators expect the typed struct to be the source of
truth, this change converts between the dynamic representation and the
typed representation on mutator entry and exit. Existing mutators can
continue to modify the typed representation and these modifications are
reflected in the dynamic representation (see `MarkMutatorEntry` and
`MarkMutatorExit` in `bundle/config/root.go`).
Required changes included in this change:
* The existing interpolation package is removed in favor of
`libs/dyn/dynvar`.
* Functionality to merge job clusters, job tasks, and pipeline clusters
are now all broken out into their own mutators.
To be implemented later:
* Allow variable references for non-string types.
* Surface diagnostics about the configuration provided by the user in
the validation output.
* Some mutators use a resource's configuration file path to resolve
related relative paths. These depend on `bundle/config/paths.Path` being
set and populated through `ConfigureConfigFilePath`. Instead, they
should interact with the dynamically typed configuration directly. Doing
this also unlocks being able to differentiate different base paths used
within a job (e.g. a task override with a relative path defined in a
directory other than the base job).
## Tests
* Existing unit tests pass (some have been modified to accommodate)
* Integration tests pass
## Changes
Some test call sites called directly into the mutator's `Apply` function
instead of `bundle.Apply`. Calling into `bundle.Apply` is preferred
because that's where we can run pre/post logic common across all
mutators.
## Tests
Pass.
## Changes
All calls to apply a mutator must go through `bundle.Apply`. This
conflicts with the existing use of the variable `bundle`. This change
un-aliases the variable from the package name by renaming all variables
to `b`.
## Tests
Pass.
## Changes
@pietern this addresses a comment from you on a recently merged PR. It
also updates settings.json based on the settings VS Code adds as soon as
you edit a notebook.
This implements the "development run" functionality that we desire for DABs in the workspace / IDE.
## bundle.yml changes
In bundle.yml, there should be a "dev" environment that is marked as
`mode: debug`:
```
environments:
dev:
default: true
mode: development # future accepted values might include pull_request, production
```
Setting `mode` to `development` indicates that this environment is used
just for running things for development. This results in several changes
to deployed assets:
* All assets will get '[dev]' in their name and will get a 'dev' tag
* All assets will be hidden from the list of assets (future work; e.g.
for jobs we would have a special job_type that hides it from the list)
* All deployed assets will be ephemeral (future work, we need some form
of garbage collection)
* Pipelines will be marked as 'development: true'
* Jobs can run on development compute through the `--compute` parameter
in the CLI
* Jobs get their schedule / triggers paused
* Jobs get concurrent runs (it's really annoying if your runs get
skipped because the last run was still in progress)
Other accepted values for `mode` are `default` (which does nothing) and
`pull-request` (which is reserved for future use).
## CLI changes
To run a single job called "shark_sighting" on existing compute, use the
following commands:
```
$ databricks bundle deploy --compute 0617-201942-9yd9g8ix
$ databricks bundle run shark_sighting
```
which would deploy and run a job called "[dev] shark_sightings" on the
compute provided. Note that `--compute` is not accepted in production
environments, so we show an error if `mode: development` is not used.
The `run --deploy` command offers a convenient shorthand for the common
combination of deploying & running:
```
$ export DATABRICKS_COMPUTE=0617-201942-9yd9g8ix
$ bundle run --deploy shark_sightings
```
The `--deploy` addition isn't really essential and I welcome feedback 🤔
I played with the idea of a "debug" or "dev" command but that seemed to
only make the option space even broader for users. The above could work
well with an IDE or workspace that automatically sets the target
compute.
One more thing I added is`run --no-wait` can now be used to run
something without waiting for it to be completed (useful for IDE-like
environments that can display progress themselves).
```
$ bundle run --deploy shark_sightings --no-wait
```
Pieter Noordhuis
Lennart Kats (databricks)