Merge pull request #33 from wpbonelli/lark

dfn parser working, still need to handle records and lists as nested contexts

flopy4/io/lark.py

@@ -0,0 +1,25 @@
+import numpy as np
+
+
+def parse_word(self, w):
+    (w,) = w
+    return str(w)
+
+
+def parse_string(self, s):
+    return " ".join(s)
+
+
+def parse_int(self, i):
+    (i,) = i
+    return int(i)
+
+
+def parse_float(self, f):
+    (f,) = f
+    return float(f)
+
+
+def parse_array(self, a):
+    (a,) = a
+    return np.array(a)

flopy4/mf6/io/converter.py

@@ -0,0 +1,3 @@
+def make_converter():
+    TODO
+    pass

flopy4/mf6/io/parser.py

@@ -5,20 +5,28 @@
 
 MF6_GRAMMAR = r"""
 // component
-component: _NL* (block _NL+)* _NL*
+component: _NL* (block _NL+)+ _NL*
 
 // block
-block: _paramblock | _listblock
-_paramblock: _BEGIN paramblock _NL params _END paramblock
+block: _dictblock | _listblock
+_dictblock: _BEGIN dictblock _NL dict _END dictblock
 _listblock: _BEGIN listblock _NL list _END listblock
-paramblock: PARAMBLOCK
+dictblock: DICTBLOCK
 listblock: LISTBLOCK [_blockindex]
 _blockindex: INT
 _BEGIN: "begin"i
 _END: "end"i
 
+// dict
+dict.1: (param _NL)*
+
+// list adapted from https://github.com/lark-parser/lark/blob/master/examples/composition/csv.lark
+// negative priority for records because the pattern is so indiscriminate
+list.-1: record*
+record.-1: _record+ _NL
+_record: scalar
+
 // parameter
-params.1: (param _NL)*
 param: key | _pair
 _pair: key value
 key: PARAM
@@ -54,12 +62,6 @@
 factor: "FACTOR" NUMBER
 iprn: "IPRN" INT
 
-// list adapted from https://github.com/lark-parser/lark/blob/master/examples/composition/csv.lark
-// negative priority for records because the pattern is so indiscriminate
-list.-1: record*
-record.-1: _record+ _NL
-_record: scalar
-
 // newline
 _NL: /(\r?\n[\t ]*)+/
 
@@ -80,7 +82,7 @@
 
 def make_parser(
     params: Iterable[str],
-    param_blocks: Iterable[str],
+    dict_blocks: Iterable[str],
     list_blocks: Iterable[str],
 ):
     """
@@ -92,18 +94,18 @@ def make_parser(
     We specify blocks containing parameters separately from blocks
     that contain a list. These must be handled separately because
     the pattern for list elements (records) casts a wider net than
-    the pattern for parameters, causing parameter blocks to parse
-    as lists otherwise.
+    the pattern for parameters, which can cause a dictionary block
+    of named parameters to parse as a block with a list of records.
 
     """
     params = "|".join(['"' + n + '"i' for n in params])
-    param_blocks = "|".join(['"' + n + '"i' for n in param_blocks])
+    dict_blocks = "|".join(['"' + n + '"i' for n in dict_blocks])
     list_blocks = "|".join(['"' + n + '"i' for n in list_blocks])
     grammar = linesep.join(
         [
             MF6_GRAMMAR,
             f"PARAM: ({params})",
-            f"PARAMBLOCK: ({param_blocks})",
+            f"DICTBLOCK: ({dict_blocks})",
             f"LISTBLOCK: ({list_blocks})",
         ]
     )

flopy4/mf6/io/spec/__init__.py

@@ -0,0 +1,4 @@
+__all__ = ["make_parser", "DFNTransformer"]
+
+from flopy4.mf6.io.spec.parser import make_parser
+from flopy4.mf6.io.spec.transformer import DFNTransformer

flopy4/mf6/io/spec/parser.py

@@ -0,0 +1,78 @@
+from os import linesep
+
+from lark import Lark
+
+ATTRIBUTES = [
+    "block",
+    "name",
+    "type",
+    "reader",
+    "optional",
+    "true",
+    "mf6internal",
+    "longname",
+    "description",
+    "layered",
+    "shape",
+    "valid",
+    "tagged",
+    "in_record",
+    "preserve_case",
+    "default_value",
+    "numeric_index",
+    "deprecated",
+]
+
+DFN_GRAMMAR = r"""
+// dfn
+dfn: _NL* (block _NL*)+ _NL*
+
+// block
+block: _header parameter*
+_header: _hash _dashes _headtext _dashes _NL+
+_headtext: component subcompnt blockname
+component: _word
+subcompnt: _word
+blockname: _word
+
+// parameter
+parameter.+1: _paramhead _NL (attribute _NL)*
+_paramhead: paramblock _NL paramname
+paramblock: "block" _word
+paramname: "name" _word
+
+// attribute
+attribute.-1: key value
+key: ATTRIBUTE
+value: string
+
+// string
+_word: /[a-zA-z0-9.;\(\)\-\,\\\/]+/
+string: _word+
+
+// newline
+_NL: /(\r?\n[\t ]*)+/
+
+// comment format
+_hash: /\#/
+_dashes: /[\-]+/
+
+%import common.SH_COMMENT -> COMMENT
+%import common.WORD
+%import common.WS_INLINE
+
+%ignore WS_INLINE
+"""
+"""
+EBNF description for the MODFLOW 6 definition language.
+"""
+
+
+def make_parser():
+    """
+    Create a parser for the MODFLOW 6 definition language.
+    """
+
+    attributes = "|".join(['"' + n + '"i' for n in ATTRIBUTES])
+    grammar = linesep.join([DFN_GRAMMAR, f"ATTRIBUTE: ({attributes})"])
+    return Lark(grammar, start="dfn")

flopy4/mf6/io/spec/transformer.py

@@ -0,0 +1,63 @@
+from lark import Transformer
+
+from flopy4.io.lark import parse_string
+
+
+class DFNTransformer(Transformer):
+    """
+    Transforms a parse tree for the MODFLOW 6
+    specification language into a nested AST
+    suitable for generating an object model.
+
+    Notes
+    -----
+    Rather than a flat list of parameters for each component,
+    which a subsequent step is responsible for turning into a
+    an object hierarchy, we derive the hierarchical parameter
+    structure from the DFN file and return a dict of blocks,
+    each of which is a dict of parameters.
+
+    This can be fed to a Jinja template to generate component
+    modules.
+    """
+
+    def key(self, k):
+        (k,) = k
+        return str(k).lower()
+
+    def value(self, v):
+        (v,) = v
+        return str(v)
+
+    def attribute(self, p):
+        return str(p[0]), str(p[1])
+
+    def parameter(self, p):
+        return dict(p[1:])
+
+    def paramname(self, n):
+        (n,) = n
+        return "name", str(n)
+
+    def paramblock(self, b):
+        (b,) = b
+        return "block", str(b)
+
+    def component(self, c):
+        (c,) = c
+        return "component", str(c)
+
+    def subcompnt(self, s):
+        (s,) = s
+        return "subcomponent", str(s)
+
+    def blockname(self, b):
+        (b,) = b
+        return "block", str(b)
+
+    def block(self, b):
+        params = {p["name"]: p for p in b[6:]}
+        return b[4][1], params
+
+    string = parse_string
+    dfn = dict

flopy4/mf6/io/transformer.py

@@ -3,6 +3,14 @@
 import numpy as np
 from lark import Transformer
 
+from flopy4.io.lark import (
+    parse_array,
+    parse_float,
+    parse_int,
+    parse_string,
+    parse_word,
+)
+
 
 class MF6Transformer(Transformer):
     """
@@ -25,29 +33,6 @@ def key(self, k):
         (k,) = k
         return str(k).lower()
 
-    def word(self, w):
-        (w,) = w
-        return str(w)
-
-    def path(self, p):
-        _, p = p
-        return Path(p)
-
-    def string(self, s):
-        return " ".join(s)
-
-    def int(self, i):
-        (i,) = i
-        return int(i)
-
-    def float(self, f):
-        (f,) = f
-        return float(f)
-
-    def array(self, a):
-        (a,) = a
-        return a
-
     def constantarray(self, a):
         # TODO factor out `ConstantArray`
         # array-like class from `MFArray`
@@ -65,27 +50,35 @@ def externalarray(self, a):
         # TODO
         pass
 
-    record = tuple
-    list = list
+    def path(self, p):
+        _, p = p
+        return Path(p)
 
     def param(self, p):
         k = p[0]
         v = True if len(p) == 1 else p[1]
         return k, v
 
-    params = dict
-
     def block(self, b):
         return tuple(b[:2])
 
-    def paramblock(self, bn):
-        return str(bn[0]).lower()
+    def dictblock(self, b):
+        return str(b[0]).lower()
 
-    def listblock(self, bn):
-        name = str(bn[0])
-        if len(bn) == 2:
-            index = int(bn[1])
+    def listblock(self, b):
+        name = str(b[0])
+        if len(b) == 2:
+            index = int(b[1])
             name = f"{name} {index}"
         return name.lower()
 
+    word = parse_word
+    string = parse_string
+    int = parse_int
+    float = parse_float
+    array = parse_array
+    record = tuple
+    list = list
+    dict = dict
+    params = dict
     component = dict