This moves the creation of a package database from given dalfs out of
the migrate command and into the init command. In particular, this makes
the process of creating a package database independent of the migrate
command.
It also changes the way this package database is created to be only
dependend on given dalf files.
* language: add internal command for generating source from dalf.
We add an internal command for generating DAML source code from .dalf
packages. Also adds an internal flag to tell the compiler, that it is
compiling generated source.
* Intern all strings and dotted names in the DAML-LF protobuf encoding
On one of our huge DAML code bases this has brought down the size of the
DALF by 60% and cut the memory usage during decoding by 50%.
The `@QUERY-LF` tests in the damlc integration tests check that the
interning actually works.
* Add scala decoder support for interned strings
* Memoize versionisOlderThan. Speeds up decoding by 6x.
* Properly fix the performance problem with version ordering
LanguageVersion.ordering was a `def` and not `val` causing us to
recreate the ordering tables every time.
* Clean up //daml-lf/archive:decode-test
* Use int32 for interning indices
* Adapt QUERY-LF tests
* Strip a few useless field_ prefixes
* Add test for uniqueness in interning tables
* Drop field_ prefix from jq queries
* Fix Scala encoder
* Fix bond-trading tests
* Rename decode-test to decode-tester
* Address @cocreature's comments
* daml-lf: proper versioning for reading internedStrings
* daml-lf repl: revert change
We're working on a feature to import plain DALFs without any attached
source or interface files into a DAML project. This PR provides a tool to
generate a simple DALF file for testing this feature.
* Remove polymorphic literals from library
* Add BECastNumeric and BEShiftNumeric in DAML-LF AST.
* Update rest
* Copyright header
* Remove polymorphic literals from LF conversion.
* Update tests
* Add test for forbidden polymorphic literals
* Add newline in DA.Numeric
* use MulNumeric instead of CastNumeric in DA.Internal.Prelude
After string interning we won't have direct access to string anymore but
rather need to go through the interning tables. This PR implements the
required infrastructure for that endeavour.
* Write JSON files in damlc integration test
There were quite a few occurences where I needed to look at the JSON
representation of the generated DALF in the past. Particularly, when
trying to come up with the `@QUERY-LF` pragmas for new tests.
This PR writes the JSON files to disk next to the (pretty printed)
DALF files.
* Remove useless rewriting of numbers into strings
* Refactor DAML-LF encoder/decoder in prep for string interning
Refactor the encoder/decoder such that all the functions concerned with
interning live next to each other and code will break once the types of
the protobuf messages change.
This PR is a pure refactoring and does not change any functionality.
* Fix typos
Right now, some functions are polymorphic in the monad we use. This is
absolutely unnecessary.
This PR fixes one monad and uses it all over the place. This is a pure
refactoring which doesn't change any functionality.
* Upgrade ghc-libs
* Convert template instances as type synonyms to DAML-LF
* Look for TEMPLATE_INSTANCE suffix for daml docs
* Update desugaring documentation
The compiler rewrites constraints like `Template (Foo t) => ...` into
`FooInstance t => ...`. Unfortunately, this rewriting does not yet kick in
when the constraint contains superfluous parentheses as in
`(Template (Foo t)) => ...`.
This PR fixes the problem.
Fixes#2994.
* Merge package reference detection into package id encoding
Currently, the encoding of a DAML-LF package works in two phases:
1. We compute the set of all package ids refernced by the package to encode.
2. We use this set of package ids to form a table for interning that is then
consulted during the encoding of package ids.
This PR changes it such that the interning table is computed while the
package is being encoded. New package ids are added to the table whenever
they are encountered for the first time.
This is in preparation for the general inlining of all strings, which will
use the same approach to building the interning table.
This PR is a pure refactoring and does not change any functionality.
* Addess @cocreature's comments
* Add unordered-containers as a dependency
Previously the test names consisted of the full absolute file paths
which got unreadable very quickly. Sadly the absolute file paths still
leak into the error messages since I don’t know of a nice way to get
relative paths here (just making them relative to cwd doesn’t work)
but this is still much more readable.
Previously, we sometimes missed the dlint hints because we got killed
due to a typechecking error before they were produced. Now, we
force them to be generated which should hopefully fix that.
Currently, the integration tests fail sometimes because of missing lints.
Our suspicion is that the linter gets killed when the type checker fails
before it managed to produce the lints. This change makes sure we wait for
the linter.
* Numeric implementation
* Dealing with all sorts of numeric literals.
* Fix DA.Generics
* Reduce code duplication with IF_NUMERIC
* Simplify Prelude with IF_NUMERIC
* Fix daml-lf validation for MUL_NUMERIC and DIV_NUMERIC
This is in preparation for interning all strings during encoding, which
will replace the current reader monad by a state monad.
This PR does not change or add any functionality. It is purely a
refactoring.
* Sync to f65e69951032a4c2ab308fd3c3471cd833925abf
* In the DAML parse tree, ListComps are coming through as MonadComps
* Linting smoke tests for lists
* Punctuate test header for readability
Under very specific circumstances, as presented in `TemplateSuperClassD`, the
template resugaring would fail and the compiler crash. This PR fixes the
problem.
It also adds test case `TemplateSuperClassF` for similar very specific
circumstances under which template resugaring would fail.
The filename of the dar is not something that you should rely on as
evidenced by the fact that we have a -o option to change it to
something completely different.
The issue was actually not really related to CPP but to the lack of
sandboxing on Windows. That resulted in us overwriting Prelude.hi from
the 1.dev rule with the Prelude.hi from the 1.6 rule in some cases
which caused the error we were seeing.
fixes#2983
* Use a consistant include dir for cwd
See https://github.com/digital-asset/ghcide/pull/114 for the actual
fix.
This PR just bumps ghcide and adds a regression test. I’ll change the
revision before merging, I just want to test CI for now.
fixes#2929
* Switch to proper ghcide revision
* writeIfacesAndHie no longer exists
* Add changelog entry
* Maybe I should try to compile code before committing but I don’t want to
* Fix ghcide exe
Not sure why this is failing but given that I won’t have the time to
investigate this today, feel free to merge this as a workaround for
now.
My current suspicion is that caching is broken somehow so maybe if the
nodes get recycled it will fix itself.
* Implement AnyTemplate in DAML
* Fix doctest path
* Shuffle around CPP
* Do not hide anything
* Hide it again
* Clean build
* Enable caching again
* debug windows crap
* more tests
* reinstantiate full tests
