We needed this for our generic template hacks. Since they are gone now,
I'm very much in favour of removing this extension since "undecidable"
always sounds a bit scary, even if it is only at compile time.
CHANGELOG_BEGIN
- [DAML Compiler] Don't make `UndecidableSuperClasses` a default language
extension for DAML anymore. If you really need this feature for a module,
you can reenable it using a `LANGUAGE` pragma at the top.
CHANGELOG_END
* Introduce a simpler template desugaring without support for generic templates
This adapts the LF conversion to the new template desugaring
introduced in our GHC fork. The guiding principle is that we use the
typeclasses directly to avoid generating, typechecking and converting
redundant code caused by indirections. I updated the template
desugaring documentation so that is probably a good starting point for
reviewing this.
* Address review comments
* Fix daml doc tests
* Fix data dependency tests
* Switch to new ghc-lib release
Display the actual number the user tried to use instead of 1.2345.
The logic to display the number is not perfect but definitely better than
we have now. If somebody knows a _simple_ way to do this better, please
tell me about it.
This reenables the migration checks. The migration command now only
generates stumbs for conversions instead of using the formerly (generic)
upgrade template.
We don't drop full support from the compiler yet but rather ban their use by
adding a check to the preprocessor. We'll remove the actual support as we go
along with fixing the upgrading story.
CHANGELOG_BEGIN
- [DAML Compiler] Make the experimental feature "generic templates"
unavailable. The current implementation is at odds with other, more important
language features still under development.
CHANGELOG_END
* daml-lf: rename Map to TextMap in archive proto
+ in Scala/haskell AST
* a bit more renamming
* Update compiler/daml-lf-tools/src/DA/Daml/LF/TypeChecker/Serializability.hs
Co-Authored-By: associahedron <231829+associahedron@users.noreply.github.com>
* fix test
* Apply suggestions from code review
Co-Authored-By: associahedron <231829+associahedron@users.noreply.github.com>
* Update compiler/daml-lf-ast/src/DA/Daml/LF/Ast/Base.hs
Co-Authored-By: associahedron <231829+associahedron@users.noreply.github.com>
* Set Sdk-Version in DAR Manifest to compiler SDK version.
CHANGELOG_BEGIN
- [DAML Compiler] Bugfix: The Sdk-Version field in a DAR manifest file
now matches the SDK version of the compiler, not the sdk-version field
from daml.yaml. These are usually the same, but they could be different
if you set the DAML_SDK_VERSION environment variable before running
``daml init`` or ``daml build``.
CHANGELOG_END
* Fallback to daml.yaml if env var not set
* add yaml dependency
* Always require sdk-version, and emit warning on mismatch with env var
* More explicit about where override comes from
* Add packaging regression test
* generate choice instances for cross sdk dependencies
* add line to test cross-sdk exercise of choice
* revert to naming choice in GenerateSimpleDalf as "NotChoice"
* remove filter/special-handlimg for "Archive" choice
* dont assume a tycon with the same name as a choice; instead use type from chcArgBinder
* error if cant find the template datatype def
* naming
* avoid intercalate followed by re-split on "." in module name
* GenerateSimpleDalf: have two choices: one of record type, one of variant type
* Update Upgrade.hs
* added some linebreaks, spaces for readability
* address cosmetic comments
* use Sdk. prefix as necessary so tests work
* mollify hlint
* fix: daml-assistant/integration-tests
* language: make only needed packages visible for cross sdk imports
This gives us more control over what packages are visible in the package
database during cross sdk imports and makes sure we're not accidentally
picking up the wrong one.
* Update compiler/damlc/lib/DA/Cli/Damlc/Packaging.hs
Co-Authored-By: Moritz Kiefer <moritz.kiefer@purelyfunctional.org>
* deal with evil dalf imports
* Move packaging tests to a separate test suite
Originally, we had these as part of the integration test suite since
`bazel run damlc build` couldn’t locate `ghc-pkg` but that has been
fixed for a while.
Moving it to a separate test suite speeds things up as the integration
tests are rerun very often and also makes development much more
convenient since the new test suite supports `-p` properly to filter
to specific tests.
For now, I’ve left the upgrading tests as part of the integration
tests. I expect that we probably want to factor those out to another
test suite as well.
* fix warnings
It has grown large enough that by now it definitely deserves to be in
its own module.
This PR does not change any of the actual code, it simply moves it around.
* language: fix for tuple types in source generation
We correct the source generation from dalfs containing TupleN data
types. A test is added to the integration tests of daml assistant and
the `generate-src` command is improved to make this test possible.
* check that tuple type comes from DA.Types
* check that compilation succeeded
* pattern match golf
* check for tuples between 2 and 20
This change allows to only import `module A` from the instances package
if data dependencies are present instead of `module A` and `module
AInstances`.
* Add toAnyContractKey and fromAnyContractKey
This is necessary to add exerciseByKey to DAML triggers.
* Switch to proper ghc-lib release
* Remove unnecessary filter
* Bump timeout because macos is terrible
* bazel fmt
I didn't really read the code around the FIXME. After spending a few moments
on it, I'm now convinces the current behaviour is the right behaviour. I
can't think of a short way to explain it in a comment without inflicting my
my previous confusion on others. Thus, I just remove the FIXME.
* No longer depend on "@bazel_tools//tools/jdk:jar"
To avoid the following warnings
```
WARNING: /home/aj/.cache/bazel/_bazel_aj/c1e06e2358666d118d0ae50e2d32c25d/external/bazel_tools/tools/jdk/BUILD:124:1: in alias rule @bazel_tools//tools/jdk🫙 target '@bazel_tools//tools/jdk:jar' depends on deprecated target '@local_jdk//:jar': Don't depend on targets in the JDK workspace; use @bazel_tools//tools/jdk:current_java_runtime instead (see https://github.com/bazelbuild/bazel/issues/5594)
```
* Targets and files should not share names
To avoid the warning
```
WARNING: /home/aj/tweag.io/da/da-master/compiler/damlc/tests/BUILD.bazel:316:1: target 'simple-dalf.dalf' is both a rule and a file; please choose another name for the rule
```
When buidling simple project that has our favourite large project as a
dependency, this decreased
- total allocations from 63GB to 57GB
- run time from 34.0s to 31.5s
* language: introduce data-imports
Right now the user experience for importing dalfs and dars from
different sdks is quiet confusing. This PR tries to solve this. We add
an additional field `data-imports` to daml.yaml. These imports can come
from different SDK's and we will generate interface files containing the
data types and their Template instances.
This also simplifies the migration command, as it now always imports the
respective packages as `data-imports`.
Given that I initialy messed this up during development (that’s where
the `reverse` in setting up the typecheck env comes from), it seems
like a good idea to have a test for this :)
* Add support for on-disk incremental builds in damlc build
* Normalise file paths of internal modules because Windows
* stop stealing my $s hlint
* Apparently jars are also called exe
* Address review comments
* Bump to proper ghcide revision
* language: cross sdk dalf/dar imports
The final piece for cross sdk imports. With this PR we can import the
data types of packages and dalfs that were created with different sdks.
This is done by generating interface files from dalfs and an 'instances'
package that contains the template instance definitions of template data
types. The instances itself are defined via the `external` keyword,
which is inlined to proper daml-lf instance definitions given in the
respective dalf package.
We test that cross sdk imports work by importing the `simple-dalf` in
the daml-assistant integation tests and running a scenario.
* Allow data A = A by prepending DamlEnum$ to type name.
* Single con enums basically work.
* Fix export lists for single constructor enum types.
* Revert "Fix export lists for single constructor enum types."
This reverts commit 7475a3dfbe3531d3ef62fdbcfe64c01a9e22d7af.
* Switch to a "stupid theta" approach
* Clean up enum type preprocessor
* Run enum preprocessor on generated code.
* Add daml-docs golden test for single constructor enums
* s/genPreprocessor/generatedPreprocessor/g
* Update copyright header
* Update release notes
* Remove unnecessary OverloadedStrings
Currently we hide modules, for which we don't define an alias, when we
specify one alias for *one* module. With this change, all modules
without aliases are still exposed.