Age | Commit message (Collapse) | Author |
|
* ccwl/ccwl.scm (<step>): Wrap constructor make-step.
[scattered-inputs, scatter-method]: New fields.
(collect-scatter-step): New function.
(collect-steps): Implement scatter.
(key->output): If step scatters, convert output to an array type.
(workflow): Add syntax to initialize scattered-inputs and
scatter-method fields of <step> object.
* ccwl/cwl.scm (workflow->cwl-scm): Add ScatterFeatureRequirement if
workflow contains scattering steps. Serialize scattered-inputs and
scatter-method.
* doc/ccwl.skb (Cookbook)[Scatter/gather]: New section.
|
|
* ccwl/ccwl.scm (key->output): Clarify in the docstring that this is
meant for workflow outputs only.
|
|
This conversation should be in ccwl.scm, not in cwl.scm. cwl.scm
should only deal with CWL serialization concerns.
* ccwl/cwl.scm (output->cwl-scm): Move conversion to ...
* ccwl/ccwl.scm (key->output): ... here.
(<output>)[type]: Add setter.
|
|
It is easier (no need to juggle complex with-syntax constructs) to do
more complicated operations outside of the scope of syntax.
* ccwl/ccwl.scm (key->output): Build output from key outside of
syntax.
|
|
* ccwl/ccwl.scm (<array-type>): New type.
(memoize, make-array-type, construct-type-syntax): New functions.
(input, output): Use construct-type-syntax.
* ccwl/cwl.scm (output->cwl-scm, input->cwl-scm): Render array types.
|
|
* ccwl/cwl.scm (output->cwl-scm): Handle both command and workflow
outputs. Accept #:workflow? argument to indicate workflow outputs.
(workflow->cwl-scm): Use output->cwl-scm.
|
|
* ccwl/cwl.scm (input->cwl-scm): New function.
(workflow->cwl-scm, command->cwl-scm): Use input->cwl-scm.
|
|
* ccwl/ccwl.scm (input=?): Delete function.
|
|
* ccwl/ui.scm (source-in-context): Do not attempt to display the line
preceding the 0th.
* tests/ui.scm ("display source in context on 0th line"): New line.
|
|
* tests/ui.scm: New file.
|
|
* build-aux/test-driver.scm (color): Use colorize-string from (term
ansi-color).
* ccwl/ui.scm (report-formatted-message, source-in-context,
report-ccwl-violation): Use colorize-string from (term ansi-color).
(color, bold, red, magenta): Delete functions.
|
|
* ccwl/ui.scm (source-in-context): Accept port, not filename.
(report-ccwl-violation): Pass port, not filename.
|
|
* ccwl/ccwl.scm (ensure-yaml-serializable): New function.
(input, output, command): Use ensure-yaml-serializable.
* tests/ccwl.scm ("inputs with #:other parameters that fail to
evaluate must raise a &ccwl-violation condition", "outputs with
#:other parameters that fail to evaluate must raise a &ccwl-violation
condition", "commands with #:other parameters that fail to evaluate
must raise a &ccwl-violation condition"): New tests.
|
|
Perhaps it fits in better with the spirit of ccwl to not evaluate
#:other parameters and instead allow only YAML serializable
trees. This saves users from quoting #:other parameters, of
course. But it also permits more thorough analysis and clearer error
messages. Besides, there is not much reason to allow general
expressions in #:other parameters. Such dynamism is not commonly
required in ccwl or CWL.
* ccwl/ccwl.scm (input, output, command): Restrict #:other parameter
to YAML serializable trees that are not evaluated.
|
|
This file was required by the previous commit, but I forgot to commit
it!
* doc/staging-input-files.scm: New file.
|
|
* ccwl/ccwl.scm (command): Add "Invalid parameter" to #:stderr and
#:stdout error messages.
* tests/ccwl.scm ("commands with non-string #:stderr parameters must
raise a &ccwl-violation condition", "commands with non-string #:stdout
parameters must raise a &ccwl-violation condition"): Adjust tests.
|
|
* ccwl/ccwl.scm (<input>)[stage?]: New field.
* ccwl/ccwl.scm (input): Add #:stage argument.
* ccwl/cwl.scm (command->cwl-scm): Serialize requirements for staged
inputs.
* doc/ccwl.skb (Cookbook)[Stage input files]: New section.
* tests/ccwl.scm ("inputs with an invalid #:stage? parameter must
raise a &ccwl-violation condition"): New test.
|
|
* ccwl/ccwl.scm (<command>)[requirements]: New field.
(command): Add #:requirements argument.
* ccwl/cwl.scm (command->cwl-scm): Serialize requirements.
|
|
Support arbitrary expressions, that are actually evaluated, in #:other
arguments. This means, among other things, that users will have to
quote them correctly. While this complicates matters for novice users
who may not be familiar with scheme, it is the most general and most
flexible solution. In this particular case, we value flexibility over
ease of use since #:other is an escape hatch and if users have to use
it, ease of use has already been lost.
* ccwl/ccwl.scm (input, output, command): Make #:other an actually
evaluated unary argument.
|
|
* ccwl/ccwl.scm (run-args): Error out on prefixes that are not
strings.
* tests/ccwl.scm ("command definitions with non-string prefixes in
prefixed inputs must raise a &ccwl-violation condition"): New test.
|
|
* ccwl/ccwl.scm (run-args): Flatten prefixed string arguments in
command definitions.
* tests/ccwl.scm ("tolerate prefixed string arguments in command
definitions"): New test.
|
|
* ccwl/ccwl.scm (command): Move run arguments construction to ...
(run-args): ... new function.
|
|
* ccwl/ccwl.scm (command): Raise &ccwl-violation on invalid command
#:run arguments.
* tests/ccwl.scm ("command definitions with invalid #:run arguments
must raise a &ccwl-violation condition"): New test.
|
|
* ccwl/ccwl.scm (command): Check if inputs used in #:run arguments of
command definitions are actually defined.
* tests/ccwl.scm ("command definitions with undefined inputs in their
#:run arguments must raise a &ccwl-violation condition", "command
definitions with undefined prefix inputs in their #:run arguments must
raise a &ccwl-violation condition"): New tests.
|
|
* tests/ccwl.scm ("allow literal as arguments"): Rename to "allow
literal strings as arguments".
(print-int): New variable.
("allow literal ints as arguments"): New test.
|
|
#:stderr and #:stdout, especially #:stdout, are commonly
required. They ought to be first class parameters and not tucked away
into #:other.
* ccwl/ccwl.scm (<command>)[stderr, stdout]: New fields.
* ccwl/ccwl.scm (command): Accept #:stderr and #:stdout as first class
parameters.
* ccwl/cwl.scm (command->cwl-scm): Serialize stderr and stdout fields.
* doc/capture-stdout.scm (print),
doc/decompress-compile-run.scm (run), doc/checksum.scm (md5sum,
sha1sum, sha256sum), doc/spell-check.scm (find-misspellings): Capture
stdout in file.
* doc/checksum.scm, doc/decompress-compile-run.scm:
* doc/ccwl.skb (Tutorial)[Capturing the standard output stream of a
command]: Document #:stdout first class parameter.
* doc/ccwl.skb (Tutorial)[Workflow with multiple steps]: Capture
stdout in explicitly named files.
* tests/ccwl.scm ("commands with non-string #:stderr parameters must
raise a &ccwl-violation condition", "commands with non-string #:stdout
parameters must raise a &ccwl-violation condition"): New tests.
|
|
* ccwl/ccwl.scm (workflow): Detect multiple expressions in workflow
body and raise an exception.
* tests/ccwl.scm ("unrecognized workflow syntaxes must raise a
&ccwl-violation condition"): Test for the formatted message too.
("multiple expressions in workflow body must raise a &ccwl-violation
condition"): New test.
|
|
* ccwl/ccwl.scm (workflow): Raise &ccwl-violation when workflow syntax
is not recognized.
* tests/ccwl.scm ("unrecognized workflow syntaxes must raise a
&ccwl-violation condition"): New test.
|
|
* tests/ccwl.scm ("step supplied with an unknown key must raise a
&ccwl-violation condition"): New test.
|
|
* ccwl/ccwl.scm (collect-steps): Convert syntax to datum before
checking if it is a symbol.
|
|
This seems like functionality that will be hard to support and will
almost never be used in a real-world workflow. This feature is not
even advertised in the manual. It is best to silently deprecate this
feature now.
* ccwl/ccwl.scm (collect-steps): Deprecate auto-connection of single
input commands.
|
|
* ccwl/ccwl.scm (<input>)[set-input-default]: New setter.
* ccwl/ccwl.scm (<command>)[set-command-inputs]: Add setter.
* ccwl/ccwl.scm (apply-partially): New function.
(collect-steps): Support literal strings as arguments.
* tests/ccwl.scm ("allow literal strings as arguments"): New test.
* .dir-locals.el (scheme-mode): Indent set-command-inputs and
set-input-default.
|
|
Many tests that use guard to catch and ensure that certain conditions
are raised actually leak and let through cases where the expressions
result in a truthy value.
* tests/ccwl.scm ("input, when passed more than one positional
argument, must raise a &ccwl-violation condition", "input, when passed
an unrecognized keyword, must raise a &ccwl-violation condition",
"input, when passed multiple arguments to a unary keyword, must raise
a &ccwl-violation condition", "output, when passed more than one
positional argument, must raise a &ccwl-violation condition", "output,
when passed an unrecognized keyword, must raise a &ccwl-violation
condition", "output, when passed multiple arguments to a unary
keyword, must raise a &ccwl-violation condition", "command, when
passed positional arguments, must raise a &ccwl-violation condition",
"command, when passed an unrecognized keyword, must raise a
&ccwl-violation condition", "command, when passed multiple arguments
to a unary keyword, must raise a &ccwl-violation condition"): Catch
expressions that don't raise exceptions.
|
|
* ccwl/ccwl.scm (collect-steps): Error out if steps with expressions
that evaluate to commands have no identifier.
* tests/ccwl.scm ("step with expression that evaluates to a command
but without a step identifier must raise a &ccwl-violation
condition"): New test.
|
|
eval is not just more portable but also more flexible. It lets us
allow steps with expressions that evaluate to commands. Sadly, we do
not yet properly support commands defined in lexical scope.
* ccwl/ccwl.scm (function-object): Use eval instead of module
introspection.
* tests/ccwl.scm ("allow steps with expressions that evaluate to
commands"): New test.
|
|
* ccwl/ccwl.scm (collect-steps): Comment on allowing default parameter
values. Remove TODO note.
|
|
* ccwl/ccwl.scm (function-inputs): New function.
(function-input-keys): Use function-inputs.
(collect-steps): Do not error out on steps with unspecified default
arguments.
* tests/ccwl.scm (print-with-default): New variable.
("allow steps with unspecified default arguments"): New test.
|
|
* scripts/ccwl (main): Raise exception when file returns neither
workflow nor command.
|
|
* scripts/ccwl (main): Replace workflow->cwl and workflow->dot with
workflow-or-command->cwl and workflow-or-command->dot.
|
|
* ccwl/cwl.scm (workflow-or-command->dot): New public function.
|
|
* ccwl/graphviz.scm (step-node, inputs-cluster, outputs-cluster,
command->graph): New functions.
(workflow->graph): Use step-node, inputs-cluster and outputs-cluster.
(command->dot): New public function.
|
|
* ccwl/cwl.scm (workflow-or-command->cwl): New public function.
|
|
* ccwl/cwl.scm (command->cwl): New public function.
|
|
* scripts/ccwl (main): Catch and report &formatted-message conditions.
|
|
* ccwl/ui.scm (report-formatted-message): New public function.
(report-ccwl-violation): Use report-formatted-message.
|
|
* scripts/ccwl: Import (ice-9 exceptions).
(main): Do not print backtrace or request the user to report a bug for
quit exceptions.
|
|
* scripts/ccwl: Set scheme mode in Emacs.
|
|
* scripts/ccwl (main): Catch uncaught exceptions and request the user
report them to the issue tracker.
|
|
Call main from the shebang command instead of calling it
directly. This allows us to run "guile -l scripts/ccwl" without
actually running main. This seems to be important for Emacs Geiser as
it freezes up without this fix.
* scripts/ccwl: Call main from shebang. Do not call main directly.
(main): Accept args as regular argument instead of as a rest argument.
|
|
* ccwl/ccwl.scm (function-object): Add TODO note to support
commands (that is, function objects) defined in lexical scope.
|