Merge pull request #25 from wpbonelli/lark

EBNF grammar for mf6 input format. with it Lark can parse a syntax tree from the input string/file with a structure matching the input component hierarchy. this is very rough but it is working minimally. not sure if we want to go this route, just experimenting

flopy4/lark/__init__.py

@@ -0,0 +1,68 @@
+from lark import Lark
+
+MF6_GRAMMAR = r"""
+?start: _NL* _item*
+_item: (block | COMMENT) _NL+
+
+// block
+block: _begin _NL params _end 
+_begin: _BEGIN name [index]
+_end: _END name
+name: WORD
+index: INT
+_BEGIN: "begin"i
+_END: "end"i
+
+// parameter
+params: (param _NL)*
+param: _key [_value]
+_key: KEYS
+_value: NUMBER | path | string | array | list
+
+// string
+string: WORD+
+
+// file path
+path: INOUT PATH
+PATH: [_PATHSEP] (NON_SEPARATOR_STRING [_PATHSEP]) [NON_SEPARATOR_STRING]
+_PATHSEP: "/"
+INOUT: "filein"i|"fileout"i
+
+// array
+array: constantarray | internalarray | externalarray
+constantarray: "CONSTANT" NUMBER
+internalarray: "INTERNAL" [factor] [iprn] (NUMBER* [_NL])*
+externalarray: "OPEN/CLOSE" WORD [factor] ["binary"] [iprn]
+factor: "FACTOR" NUMBER
+iprn: "IPRN" INT
+
+// list (adapted from https://github.com/lark-parser/lark/blob/master/examples/composition/csv.lark)
+list: header _NL row*
+header: "#" " "? (WORD _SEPARATOR?)+
+row: (_anything _SEPARATOR?)+ _NL
+_anything: INT | WORD | NON_SEPARATOR_STRING | FLOAT | SIGNED_FLOAT
+NON_SEPARATOR_STRING: /[a-zA-z.;\\\/]+/
+_SEPARATOR: /[  ]+/
+          | "\t"
+          | ","
+
+// newline
+_NL: /(\r?\n[\t ]*)+/
+
+// parameter keys file can be generated
+// with the rest of the plugin interface
+// and maybe placed in a separate file
+KEYS: "K"|"I"|"D"|"S"|"F"|"A"
+
+%import common.SH_COMMENT -> COMMENT
+%import common.SIGNED_NUMBER -> NUMBER
+%import common.SIGNED_FLOAT
+%import common.INT
+%import common.FLOAT
+%import common.WORD
+%import common.WS_INLINE
+
+%ignore WS_INLINE
+"""
+
+MF6_PARSER = Lark(MF6_GRAMMAR, start="start")

flopy4/param.py

@@ -29,8 +29,6 @@ class MFParamSpec:
     repeating: bool = False
     tagged: bool = True
     reader: MFReader = MFReader.urword
-    # todo change to variadic tuple of str and resolve
-    # actual shape at load time from simulation context
     shape: Optional[Tuple[int]] = None
     default_value: Optional[Any] = None
 

pyproject.toml

@@ -36,10 +36,11 @@ classifiers = [
 ]
 requires-python = ">=3.9"
 dependencies = [
-    "attrs",  # todo: lower bound?
-    "cattrs",  # todo: lower bound?
-    "Jinja2>=3.0",
+    "attrs",  # todo: bounds?
+    "cattrs",  # todo: bounds?
     "flopy>=3.7.0",
+    "Jinja2>=3.0",
+    "lark",  # todo: bounds?
     "numpy>=1.20.3",
     "pandas>=2.0.0",
     "toml>=0.10",
@@ -54,6 +55,7 @@ test = [
     "flopy4[lint]",
     "coverage",
     "GitPython",
+    "interegular",
     "jupyter",
     "jupytext",
     "modflow-devtools",

test/test_lark.py

@@ -0,0 +1,42 @@
+from pprint import pprint
+
+import pytest
+from lark import Transformer
+
+from flopy4.lark import MF6_PARSER
+
+TEST_PKG = """
+BEGIN OPTIONS
+  K
+  I 1
+  D 1.0
+  S hello world
+  F FILEIN some/path
+END OPTIONS
+
+BEGIN PACKAGEDATA 1
+  A INTERNAL 1.0 2.0 3.0
+END PACKAGEDATA
+"""
+
+
+def test_parse_mf6():
+    tree = MF6_PARSER.parse(TEST_PKG)
+    # this is working, check it with:
+    # pytest test/test_lark.py::test_parse_mf6 -s
+    print(tree.pretty())
+
+
+class MF6Transformer(Transformer):
+    # TODO
+    pass
+
+
+MF6_TRANSFORMER = MF6Transformer()
+
+
+@pytest.mark.xfail
+def test_transform_mf6():
+    tree = MF6_PARSER.parse(TEST_PKG)
+    data = MF6_TRANSFORMER.transform(tree)
+    pprint(data)