I had a look at the remaining places mentioning --heuristic, --evaluator, and --landmarks. Most of them are about supporting the old syntax, or documenting that the old syntax is now deprecated:
docs/search-plugin-syntax.md: Documentation warning about deprecation
src/search/plugins/doc_printer.cc: Documentation warning about deprecation
CHANGES.md: Documentation warning about deprecation
+ one older change log entry from the time
when the option still existed
src/search/command_line.cc: rewriting old syntax into let expressions
driver/portfolio_runner.py: Support for the legacy arguments
Other than that, I only found 3 remaining cases:
misc/tests/configs.py:
The bjolp test config still uses the old "--evaluator".
We can just rewrite it in the new style.
driver/arguments.py:
One of the example configurations in the driver shows a predefinition
but marked as "(old style)" with the new style above.
We could just remove it.
docs/search-plugin-syntax.md:
This contains an example that is more complex to rewrite:
--if-unit-cost --evaluator "h1=ff()" --evaluator "h2=blind()" \
--if-non-unit-cost --evaluator "h1=cea()" --evaluator "h2=lmcut()" \
--always --search "eager_greedy([h1, h2])"
One option is
--if-unit-cost --search \
"let(h1, ff(), let(h2, blind(), eager_greedy([h1, h2])))" \
--if-non-unit-cost --search \
"let(h1, cea(), let(h2, lmcut(), eager_greedy([h1, h2])))"
but it duplicates the part that previously was written as --always.
|