micropython-stubber 1.23.1__py3-none-any.whl → 1.23.2__py3-none-any.whl

stubber/rst/rst_utils.py

@@ -1,540 +1,541 @@
-"""
-Work in Progress
-----------------
-
-Tries to determine the return type by parsing the docstring and the function signature
- - if the signature contains a return type --> <something> then that is returned
- - check a lookup dictionary of type overrides, 
-    if the functionnae is listed, then use the override
- - use re to find phrases such as:
-    - 'Returns ..... '
-    - 'Gets  ..... '
- - docstring is joined without newlines to simplify parsing
- - then parses the docstring to find references to known types and give then a rating though a hand coded model ()
- - builds a list return type candidates 
- - selects the highest ranking candidate 
- - the default Type is 'Any'
- 
-
-to do:
-
-    - regex :
-        - 'With no arguments the frequency in Hz is returned.'
-        - 'Get or set' --> indicates overloaded/optional return Union[None|...]
-        - add regex for 'Query' ` Otherwise, query current state if no argument is provided. `
-
-    - regex :
-        - 'With no arguments the frequency in Hz is returned.'
-        - 'Get or set' --> indicates overloaded/optional return Union[None|...]
-        - add regex for 'Query' ` Otherwise, query current state if no argument is provided. `
-
-    - try if an Azure Machine Learning works as well 
-        https://docs.microsoft.com/en-us/azure/machine-learning/quickstart-create-resources
-    - 
-"""
-# ref: https://regex101.com/codegen?language=python
-# https://regex101.com/r/Ni8g2z/2
-
-import re
-from typing import Dict, List, Optional, Union
-
-from loguru import logger as log
-
-from .lookup import LOOKUP_LIST, NONE_VERBS, TYPING_IMPORT
-
-# These are shown to import
-__all__ = [
-    "simple_candidates",
-    "compound_candidates",
-    "object_candidates",
-    "distill_return",
-    "return_type_from_context",
-    "_type_from_context",  # For testing only
-    "TYPING_IMPORT",
-]
-
-
-# logging
-# # log = logging.getLogger(__name__)
-
-
-# --------------------------------------
-# Confidence levels
-# these heuristics are based a significant amout of manual testing,
-# and not based on any statistical analysis
-
-C_DEFAULT = 0  # Any , the default for all
-C_NONE = 0.1 + C_DEFAULT  # better than the default Any
-C_BASE = 0.1 + C_NONE  # the Base if a return type has been found
-
-C_STR_NAMES = 0.3
-
-C_GENERIC = 0.6
-C_DICT = C_GENERIC
-C_TUPLE = C_GENERIC
-C_LIST = C_GENERIC
-C_BOOL = C_GENERIC
-C_FLOAT = C_GENERIC
-C_STR = C_GENERIC
-
-# tehere is a bit of logic - but mostly empirical
-C_NONE_RETURN = C_GENERIC
-C_OBJECTS = 0.01 + C_GENERIC
-
-C_BYTES = 0.01 + C_GENERIC
-C_BYTEARRAY = 0.03 + C_GENERIC
-C_INT = 0.03 + C_GENERIC
-C_UINT = 0.04 + C_GENERIC
-C_ITERATOR = 0.4 + C_GENERIC
-C_GENERATOR = 0.4 + C_GENERIC
-
-C_INT_SIZES = 0.5 + C_GENERIC  # better match than bytes and bytearray or object
-C_INT_LIKE = 0.5 + C_GENERIC
-
-C_LOOKUP = C_GENERIC + 1
-
-# --------------------------------------
-# Weights of the different Lookups
-WEIGHT_LOOPUPS = 3.0  # Lookup list weight factor
-WEIGHT_RETURN_VAL = 3.0  # Lookup list weight factor
-WEIGHT_RETURNS = 1.8  # for Docstring returns
-WEIGHT_GETS = 1.5  # For docstring Gets
-
-# --------------------------------------
-
-# base has a confidence that is quite low, but better than rubbish
-BASE = {"type": "Incomplete", "confidence": C_BASE, "match": None}
-
-# --------------------------------------
-# Regexes
-# --------------------------------------
-
-# all regex matches stop at end of sentence:: . ! ? : ;
-# Look for "Return Value: xxxx"
-RE_RETURN_VALUE = r"Return value\s?:\s?(?P<return>[^.!?:;]*)"
-# Look for Returns , but no 'Information'
-RE_RETURN = r"Return(?:s?,?|(?:ing)?)\s(?!information)(?P<return>[^.!?:;]*)"
-# Look for gets
-RE_GETS = r"Gets?\s(?P<return>[^.!?:;]*)"
-
-# --------------------------------------
-# Regex for Literals
-# --------------------------------------
-RE_LIT_AS_A = r"as a\s?(?P<return>[^.!?:;]*)"
-RE_LIT_SENTENCE = r"\s?(?P<return>[^.!?:;]*)"
-
-
-def dist_rate(i: int) -> float:
-    """"""
-    max_len = 150  # must occur in the first 150 chars
-    return max((max_len - i), 1) / max_len
-
-
-WORD_TERMINATORS = ".,!;:?"
-
-
-def simple_candidates(
-    type: str,
-    match_string: str,
-    keywords: List[str],
-    rate: float = 0.5,
-    exclude: Optional[List[str]] = None,
-):
-    """
-    find and rate possible types and confidence weighting for simple types.
-    Case sensitive
-    """
-    if exclude is None:
-        exclude = []
-    candidates = []
-    if not any(t in match_string for t in keywords) or any(t in match_string for t in exclude):
-        # quick bailout , there are no matches, or there is an exclude
-        return []
-
-    #  word matching
-    match_words = [w.strip(WORD_TERMINATORS) for w in match_string.split()]
-    #  kw =  single word -
-    for kw in keywords:
-        i = match_string.find(kw)
-        if " " not in kw and kw not in match_words or " " in kw and i < 0:
-            continue
-        # Assume unsigned are int
-        result = BASE.copy()
-        result["type"] = type
-        result["confidence"] = rate * dist_rate(i)  # OK
-        log.trace(f" - found '{kw}' at position {i} with rating {dist_rate(i)}")
-        candidates.append(result)
-    return candidates
-
-
-def compound_candidates(
-    type: str,
-    match_string: str,
-    keywords: List[str],
-    rate: float = 0.85,
-    exclude: Optional[List[str]] = None,
-):
-    """
-    find and rate possible types and confidence weighting for compound types that can have a subscription.
-    Case sensitive
-    """
-    if exclude is None:
-        exclude = []
-    candidates = []
-    if not any(t in match_string for t in keywords) or any(t in match_string for t in exclude):
-        # quick bailout , there are no matches, or there is an exclude
-        return []
-
-    #  word matching
-    match_words = [w.strip(WORD_TERMINATORS) for w in match_string.split()]
-    #  kw =  single word -
-    for kw in keywords:
-        i = match_string.find(kw)
-        if " " not in kw and kw not in match_words or " " in kw and i < 0:
-            continue
-        # List / Dict / Generator of Any / Tuple /
-        sub = None
-        result = BASE.copy()
-        confidence = rate
-        for element in ("tuple", "string", "unsigned", "int"):
-            if element in match_string.casefold():
-                j = match_string.find(element)
-                if i == j:
-                    # do not match on the same main and sub
-                    continue
-                confidence += 0.10  # boost as we have a subtype
-                if element == "string":
-                    sub = "str"
-                    break
-                elif element == "tuple":
-                    sub = "Tuple"
-                    break
-                elif element == "unsigned":
-                    sub = "int"
-                    break
-                else:
-                    sub = element
-        result["type"] = f"{type}[{sub}]" if sub else f"{type}"
-        confidence = confidence * dist_rate(i)  # distance weighting
-        result["confidence"] = confidence
-        log.trace(
-            f" - found '{kw}' at position {i} with confidence {confidence} rating {dist_rate(i)}"
-        )
-
-        candidates.append(result)
-    return candidates
-
-
-def object_candidates(match_string: str, rate: float = 0.81, exclude: Optional[List[str]] = None):
-    """
-    find and rate possible types and confidence weighting for Object types.
-    Case sensitive
-    Exclude defaults to ["IRQ"]
-    """
-    # defaults
-    if exclude is None:
-        exclude = ["IRQ"]
-    candidates = []
-    keywords = [
-        "Object",
-        "object",
-    ]  # Q&D
-
-    if not any(t in match_string for t in keywords) or any(t in match_string for t in exclude):
-        # quick bailout , there are no matches, or there is an exclude
-        return []
-    for kw in keywords:
-        i = match_string.find(kw)
-        if i < 0:
-            continue
-        # List / Dict / Generator of Any / Tuple /
-        confidence = rate
-
-        # did the word actually occur, or is it just a partial
-        words = match_string.split(" ")  # Return <multiple words object>
-        if kw in words:
-            pos = words.index(kw)
-            obj = "Incomplete" if pos == 0 else words[pos - 1]
-            if obj in ("stream-like", "file"):
-                obj = "IO"  # needs from typing import IO
-            elif obj == "callback":
-                obj = "Callable[..., Incomplete]"  # requires additional 'from typing import Callable'
-            else:
-                # clean
-                obj = re.sub(r"[^a-z.A-Z0-9]", "", obj)
-            result = BASE.copy()
-            result["type"] = obj
-            if obj in ["an", "any"]:  # "Return an / any object"
-                result["type"] = "Incomplete"
-                confidence += 0.10  # abstract , but very good
-            elif obj[0].islower():
-                confidence -= 0.20  # not so good
-            result["confidence"] = confidence * dist_rate(i)
-            candidates.append(result)
-    return candidates
-
-
-def has_none_verb(docstr: str) -> List:
-    "returns a None result if the docstring starts with a verb that indicates None"
-    docstr = docstr.strip().casefold()
-    if not any(docstr.startswith(kw.casefold()) for kw in NONE_VERBS):
-        return []
-    result = BASE.copy()
-    result["type"] = "None"
-    result["confidence"] = C_NONE  # better than the default Any
-    return [result]
-
-
-def distill_return(return_text: str) -> List[Dict]:
-    """Find return type and confidence.
-    Returns a list of possible types and confidence weighting.
-    {
-
-        type :str               # the return type
-        confidence: float       # the confidence between 0.0 and 1
-        match: Optional[str]    # for debugging : the reason the match was made
-
-    }
-
-    """
-    candidates = [BASE]  # Default to the base , which is 'Any'
-
-    # clean up match_string
-    match_string = return_text.strip().rstrip(".")
-    match_string = match_string.replace("`", "")
-
-    candidates += compound_candidates("Generator", match_string, ["generator"], C_GENERATOR)
-    candidates += compound_candidates("Iterator", match_string, ["iterator"], C_ITERATOR)
-    candidates += compound_candidates(
-        "List", match_string, ["a list of", "list of", "an array"], C_LIST
-    )
-
-    candidates += simple_candidates(
-        "Dict", match_string, ["a dictionary", "dict", "Dictionary"], C_DICT
-    )
-    candidates += simple_candidates(
-        "Tuple",
-        match_string,
-        [
-            "tuple",
-            "a pair",
-            "1-tuple",
-            "2-tuple",
-            "3-tuple",
-            "4-tuple",
-            "5-tuple",
-            "6-tuple",
-            "7-tuple",
-            "8-tuple",
-            "9-tuple",
-        ],
-        C_TUPLE,
-    )
-
-    candidates += simple_candidates(
-        "int", match_string, ["unsigned integer", "unsigned int", "unsigned"], C_UINT
-    )
-
-    candidates += simple_candidates(
-        "int",
-        match_string,
-        [
-            "number",
-            "integer",
-            "count",
-            "int",
-            "0 or 1",
-        ],
-        C_INT,
-    )
-
-    # good but nor perfect indicators of integers
-    # better match than bytes and bytearray or object
-    candidates += simple_candidates(
-        "int",
-        match_string,
-        [
-            "length",
-            "total size",
-            "size of",
-            "the index",
-            "number of",
-            "address of",
-            "the duration",
-        ],
-        C_INT_SIZES,
-    )
-
-    candidates += simple_candidates("int", match_string, [], C_INT_SIZES)
-
-    # Assume numbers are signed int
-    candidates += simple_candidates(
-        "int",
-        match_string,
-        [
-            "index",
-            "**signed** value",
-            "seconds",
-            "nanoseconds",
-            "millisecond",
-            "offset",
-        ],
-        C_INT_LIKE,
-    )
-
-    # better match than bytes
-    candidates += simple_candidates("bytearray", match_string, ["bytearray"], C_BYTEARRAY)
-
-    # OK, better than just string
-    candidates += simple_candidates("bytes", match_string, ["bytes", "byte string"], C_BYTES)
-
-    candidates += simple_candidates(
-        "bool", match_string, ["boolean", "bool", "True", "False"], C_BOOL
-    )
-    candidates += simple_candidates(
-        "float",
-        match_string,
-        [
-            "float",
-            "logarithm",
-            "sine",
-            "cosine",
-            "tangent",
-            "exponential",
-            "complex number",
-            "phase",
-            "ratio of",
-        ],
-        C_FLOAT,
-    )
-
-    candidates += simple_candidates(
-        "str", match_string, ["string", "(sub)string", "sub-string", "substring"], C_STR
-    )
-
-    candidates += simple_candidates("str", match_string, ["name", "names"], C_STR_NAMES)
-    ## "? contains 'None if there is no'  --> Union[Null, xxx]"
-    candidates += simple_candidates(
-        "None",
-        match_string,
-        ["``None``", "None"],
-        C_NONE_RETURN,
-        exclude=["previous value", "if there is no"],
-    )
-
-    candidates += object_candidates(match_string, C_OBJECTS)
-
-    return candidates
-
-
-def return_type_from_context(
-    *, docstring: Union[str, List[str]], signature: str, module: str, literal: bool = False
-):
-    try:
-        return str(
-            _type_from_context(
-                module=module, signature=signature, docstring=docstring, literal=literal
-            )["type"]
-        )
-    except Exception:
-        return "Incomplete"
-
-
-def _type_from_context(
-    *, docstring: Union[str, List[str]], signature: str, module: str, literal: bool = False
-):  # -> Dict[str , Union[str,float]]:
-    """Determine the return type of a function or method based on:
-     - the function signature
-     - the terminology used in the docstring
-
-    Logic:
-    - if the signature contains a return type --> <something> then that is returned
-    - use re to find phrases such as:
-
-        - 'Returns ..... '
-        - 'Gets  ..... '
-
-    - docstring is joined without newlines to simplify parsing
-    - then parses the docstring to find references to known types and give then a rating though a hand coded model ()
-    - builds a list return type candidates
-    - selects the highest ranking candidate
-    - the default Type is 'Any'
-    """
-
-    if isinstance(docstring, list):
-        # join with space to avoid ending at a newline
-        docstring = " ".join(docstring)
-
-    # give the regex that searches for returns a 0.2 boost as that is bound to be more relevant
-
-    weighted_regex = (
-        [
-            (RE_LIT_AS_A, 1.0),
-            (RE_LIT_SENTENCE, 2.0),
-        ]
-        if literal
-        else [
-            (RE_RETURN_VALUE, WEIGHT_RETURN_VAL),
-            (RE_RETURN, WEIGHT_RETURNS),
-            (RE_GETS, WEIGHT_GETS),
-            #       (reads_regex, 1.0),
-        ]
-    )
-    # only the function name without the leading module
-    function_re = re.compile(r"[\w|.]+(?=\()")
-
-    # matches: List[re.Match] = []
-    candidates: List[Dict] = [{"match": "default", "type": "Incomplete", "confidence": 0}]
-
-    # if the signature contains a return type , then use that and do nothing else.
-    if "->" in signature:
-        sig_type = signature.split("->")[-1].strip(": ")
-        return {"type": sig_type, "confidence": WEIGHT_LOOPUPS, "match": signature}
-
-    # ------------------------------------------------------
-    # lookup returns that cannot be found based on the docstring from the lookup list
-    try:
-        function_name = function_re.findall(signature)[0]
-    except IndexError:
-        function_name = signature.strip().strip(":()")
-
-    function_name = ".".join((module, function_name))
-
-    if function_name in LOOKUP_LIST.keys():
-        sig_type = LOOKUP_LIST[function_name][0]
-        return {
-            "type": sig_type,
-            "confidence": C_LOOKUP * WEIGHT_LOOPUPS,
-            "match": function_name,
-        }
-    # ------------------------------------------------------
-    # parse the docstring for simple start verbs,
-    # and add them as a candidate
-    candidates += has_none_verb(docstring)
-
-    # ------------------------------------------------------
-    # parse the docstring for the regexes and weigh the results accordingly
-    for weighted in weighted_regex:
-        match_iter = re.finditer(weighted[0], docstring, re.MULTILINE | re.IGNORECASE)
-        for match in match_iter:
-            # matches.append(match)
-            distilled = distill_return(match.group("return"))
-            for item in distilled:
-                candidate = {
-                    "match": match,
-                    "type": item["type"],
-                    "confidence": item["confidence"] * weighted[1],  # add search boost
-                }
-                candidates.append(candidate)
-    # Sort
-    candidates = sorted(candidates, key=lambda x: x["confidence"], reverse=True)
-    best = candidates[0]  # best candidate
-
-    # ref: https://docs.python.org/3/library/typing.html#typing.Coroutine
-    # Coroutine[YieldType, SendType, ReturnType]
-    # todo: sanity check against actual code .....
-    if "This is a coroutine" in docstring and "Coroutine" not in str(best["type"]):  # type: ignore
-        best["type"] = f"Coroutine[{best['type']}, Any, Any]"
-
-    # return the best candidate, or Any
-    return best  # best candidate
+"""
+Work in Progress
+----------------
+
+Tries to determine the return type by parsing the docstring and the function signature
+ - if the signature contains a return type --> <something> then that is returned
+ - check a lookup dictionary of type overrides, 
+    if the functionnae is listed, then use the override
+ - use re to find phrases such as:
+    - 'Returns ..... '
+    - 'Gets  ..... '
+ - docstring is joined without newlines to simplify parsing
+ - then parses the docstring to find references to known types and give then a rating though a hand coded model ()
+ - builds a list return type candidates 
+ - selects the highest ranking candidate 
+ - the default Type is 'Any'
+ 
+
+to do:
+
+    - regex :
+        - 'With no arguments the frequency in Hz is returned.'
+        - 'Get or set' --> indicates overloaded/optional return Union[None|...]
+        - add regex for 'Query' ` Otherwise, query current state if no argument is provided. `
+
+    - regex :
+        - 'With no arguments the frequency in Hz is returned.'
+        - 'Get or set' --> indicates overloaded/optional return Union[None|...]
+        - add regex for 'Query' ` Otherwise, query current state if no argument is provided. `
+
+    - try if an Azure Machine Learning works as well 
+        https://docs.microsoft.com/en-us/azure/machine-learning/quickstart-create-resources
+    - 
+"""
+
+# ref: https://regex101.com/codegen?language=python
+# https://regex101.com/r/Ni8g2z/2
+
+import re
+from typing import Dict, List, Optional, Union
+
+from mpflash.logger import log
+
+from .lookup import LOOKUP_LIST, NONE_VERBS, TYPING_IMPORT
+
+# These are shown to import
+__all__ = [
+    "simple_candidates",
+    "compound_candidates",
+    "object_candidates",
+    "distill_return",
+    "return_type_from_context",
+    "_type_from_context",  # For testing only
+    "TYPING_IMPORT",
+]
+
+
+# logging
+# # log = logging.getLogger(__name__)
+
+
+# --------------------------------------
+# Confidence levels
+# these heuristics are based a significant amout of manual testing,
+# and not based on any statistical analysis
+
+C_DEFAULT = 0  # Any , the default for all
+C_NONE = 0.1 + C_DEFAULT  # better than the default Any
+C_BASE = 0.1 + C_NONE  # the Base if a return type has been found
+
+C_STR_NAMES = 0.3
+
+C_GENERIC = 0.6
+C_DICT = C_GENERIC
+C_TUPLE = C_GENERIC
+C_LIST = C_GENERIC
+C_BOOL = C_GENERIC
+C_FLOAT = C_GENERIC
+C_STR = C_GENERIC
+
+# tehere is a bit of logic - but mostly empirical
+C_NONE_RETURN = C_GENERIC
+C_OBJECTS = 0.01 + C_GENERIC
+
+C_BYTES = 0.01 + C_GENERIC
+C_BYTEARRAY = 0.03 + C_GENERIC
+C_INT = 0.03 + C_GENERIC
+C_UINT = 0.04 + C_GENERIC
+C_ITERATOR = 0.4 + C_GENERIC
+C_GENERATOR = 0.4 + C_GENERIC
+
+C_INT_SIZES = 0.5 + C_GENERIC  # better match than bytes and bytearray or object
+C_INT_LIKE = 0.5 + C_GENERIC
+
+C_LOOKUP = C_GENERIC + 1
+
+# --------------------------------------
+# Weights of the different Lookups
+WEIGHT_LOOPUPS = 3.0  # Lookup list weight factor
+WEIGHT_RETURN_VAL = 3.0  # Lookup list weight factor
+WEIGHT_RETURNS = 1.8  # for Docstring returns
+WEIGHT_GETS = 1.5  # For docstring Gets
+
+# --------------------------------------
+
+# base has a confidence that is quite low, but better than rubbish
+BASE = {"type": "Incomplete", "confidence": C_BASE, "match": None}
+
+# --------------------------------------
+# Regexes
+# --------------------------------------
+
+# all regex matches stop at end of sentence:: . ! ? : ;
+# Look for "Return Value: xxxx"
+RE_RETURN_VALUE = r"Return value\s?:\s?(?P<return>[^.!?:;]*)"
+# Look for Returns , but no 'Information'
+RE_RETURN = r"Return(?:s?,?|(?:ing)?)\s(?!information)(?P<return>[^.!?:;]*)"
+# Look for gets
+RE_GETS = r"Gets?\s(?P<return>[^.!?:;]*)"
+
+# --------------------------------------
+# Regex for Literals
+# --------------------------------------
+RE_LIT_AS_A = r"as a\s?(?P<return>[^.!?:;]*)"
+RE_LIT_SENTENCE = r"\s?(?P<return>[^.!?:;]*)"
+
+
+def dist_rate(i: int) -> float:
+    """"""
+    max_len = 150  # must occur in the first 150 chars
+    return max((max_len - i), 1) / max_len
+
+
+WORD_TERMINATORS = ".,!;:?"
+
+
+def simple_candidates(
+    type: str,
+    match_string: str,
+    keywords: List[str],
+    rate: float = 0.5,
+    exclude: Optional[List[str]] = None,
+):
+    """
+    find and rate possible types and confidence weighting for simple types.
+    Case sensitive
+    """
+    if exclude is None:
+        exclude = []
+    candidates = []
+    if not any(t in match_string for t in keywords) or any(t in match_string for t in exclude):
+        # quick bailout , there are no matches, or there is an exclude
+        return []
+
+    #  word matching
+    match_words = [w.strip(WORD_TERMINATORS) for w in match_string.split()]
+    #  kw =  single word -
+    for kw in keywords:
+        i = match_string.find(kw)
+        if " " not in kw and kw not in match_words or " " in kw and i < 0:
+            continue
+        # Assume unsigned are int
+        result = BASE.copy()
+        result["type"] = type
+        result["confidence"] = rate * dist_rate(i)  # OK
+        log.trace(f" - found '{kw}' at position {i} with rating {dist_rate(i)}")
+        candidates.append(result)
+    return candidates
+
+
+def compound_candidates(
+    type: str,
+    match_string: str,
+    keywords: List[str],
+    rate: float = 0.85,
+    exclude: Optional[List[str]] = None,
+):
+    """
+    find and rate possible types and confidence weighting for compound types that can have a subscription.
+    Case sensitive
+    """
+    if exclude is None:
+        exclude = []
+    candidates = []
+    if not any(t in match_string for t in keywords) or any(t in match_string for t in exclude):
+        # quick bailout , there are no matches, or there is an exclude
+        return []
+
+    #  word matching
+    match_words = [w.strip(WORD_TERMINATORS) for w in match_string.split()]
+    #  kw =  single word -
+    for kw in keywords:
+        i = match_string.find(kw)
+        if " " not in kw and kw not in match_words or " " in kw and i < 0:
+            continue
+        # List / Dict / Generator of Any / Tuple /
+        sub = None
+        result = BASE.copy()
+        confidence = rate
+        for element in ("tuple", "string", "unsigned", "int"):
+            if element in match_string.casefold():
+                j = match_string.find(element)
+                if i == j:
+                    # do not match on the same main and sub
+                    continue
+                confidence += 0.10  # boost as we have a subtype
+                if element == "string":
+                    sub = "str"
+                    break
+                elif element == "tuple":
+                    sub = "Tuple"
+                    break
+                elif element == "unsigned":
+                    sub = "int"
+                    break
+                else:
+                    sub = element
+        result["type"] = f"{type}[{sub}]" if sub else f"{type}"
+        confidence = confidence * dist_rate(i)  # distance weighting
+        result["confidence"] = confidence
+        log.trace(
+            f" - found '{kw}' at position {i} with confidence {confidence} rating {dist_rate(i)}"
+        )
+
+        candidates.append(result)
+    return candidates
+
+
+def object_candidates(match_string: str, rate: float = 0.81, exclude: Optional[List[str]] = None):
+    """
+    find and rate possible types and confidence weighting for Object types.
+    Case sensitive
+    Exclude defaults to ["IRQ"]
+    """
+    # defaults
+    if exclude is None:
+        exclude = ["IRQ"]
+    candidates = []
+    keywords = [
+        "Object",
+        "object",
+    ]  # Q&D
+
+    if not any(t in match_string for t in keywords) or any(t in match_string for t in exclude):
+        # quick bailout , there are no matches, or there is an exclude
+        return []
+    for kw in keywords:
+        i = match_string.find(kw)
+        if i < 0:
+            continue
+        # List / Dict / Generator of Any / Tuple /
+        confidence = rate
+
+        # did the word actually occur, or is it just a partial
+        words = match_string.split(" ")  # Return <multiple words object>
+        if kw in words:
+            pos = words.index(kw)
+            obj = "Incomplete" if pos == 0 else words[pos - 1]
+            if obj in ("stream-like", "file"):
+                obj = "IO"  # needs from typing import IO
+            elif obj == "callback":
+                obj = "Callable[..., Incomplete]"  # requires additional 'from typing import Callable'
+            else:
+                # clean
+                obj = re.sub(r"[^a-z.A-Z0-9]", "", obj)
+            result = BASE.copy()
+            result["type"] = obj
+            if obj in ["an", "any"]:  # "Return an / any object"
+                result["type"] = "Incomplete"
+                confidence += 0.10  # abstract , but very good
+            elif obj[0].islower():
+                confidence -= 0.20  # not so good
+            result["confidence"] = confidence * dist_rate(i)
+            candidates.append(result)
+    return candidates
+
+
+def has_none_verb(docstr: str) -> List:
+    "returns a None result if the docstring starts with a verb that indicates None"
+    docstr = docstr.strip().casefold()
+    if not any(docstr.startswith(kw.casefold()) for kw in NONE_VERBS):
+        return []
+    result = BASE.copy()
+    result["type"] = "None"
+    result["confidence"] = C_NONE  # better than the default Any
+    return [result]
+
+
+def distill_return(return_text: str) -> List[Dict]:
+    """Find return type and confidence.
+    Returns a list of possible types and confidence weighting.
+    {
+
+        type :str               # the return type
+        confidence: float       # the confidence between 0.0 and 1
+        match: Optional[str]    # for debugging : the reason the match was made
+
+    }
+
+    """
+    candidates = [BASE]  # Default to the base , which is 'Any'
+
+    # clean up match_string
+    match_string = return_text.strip().rstrip(".")
+    match_string = match_string.replace("`", "")
+
+    candidates += compound_candidates("Generator", match_string, ["generator"], C_GENERATOR)
+    candidates += compound_candidates("Iterator", match_string, ["iterator"], C_ITERATOR)
+    candidates += compound_candidates(
+        "List", match_string, ["a list of", "list of", "an array"], C_LIST
+    )
+
+    candidates += simple_candidates(
+        "Dict", match_string, ["a dictionary", "dict", "Dictionary"], C_DICT
+    )
+    candidates += simple_candidates(
+        "Tuple",
+        match_string,
+        [
+            "tuple",
+            "a pair",
+            "1-tuple",
+            "2-tuple",
+            "3-tuple",
+            "4-tuple",
+            "5-tuple",
+            "6-tuple",
+            "7-tuple",
+            "8-tuple",
+            "9-tuple",
+        ],
+        C_TUPLE,
+    )
+
+    candidates += simple_candidates(
+        "int", match_string, ["unsigned integer", "unsigned int", "unsigned"], C_UINT
+    )
+
+    candidates += simple_candidates(
+        "int",
+        match_string,
+        [
+            "number",
+            "integer",
+            "count",
+            "int",
+            "0 or 1",
+        ],
+        C_INT,
+    )
+
+    # good but nor perfect indicators of integers
+    # better match than bytes and bytearray or object
+    candidates += simple_candidates(
+        "int",
+        match_string,
+        [
+            "length",
+            "total size",
+            "size of",
+            "the index",
+            "number of",
+            "address of",
+            "the duration",
+        ],
+        C_INT_SIZES,
+    )
+
+    candidates += simple_candidates("int", match_string, [], C_INT_SIZES)
+
+    # Assume numbers are signed int
+    candidates += simple_candidates(
+        "int",
+        match_string,
+        [
+            "index",
+            "**signed** value",
+            "seconds",
+            "nanoseconds",
+            "millisecond",
+            "offset",
+        ],
+        C_INT_LIKE,
+    )
+
+    # better match than bytes
+    candidates += simple_candidates("bytearray", match_string, ["bytearray"], C_BYTEARRAY)
+
+    # OK, better than just string
+    candidates += simple_candidates("bytes", match_string, ["bytes", "byte string"], C_BYTES)
+
+    candidates += simple_candidates(
+        "bool", match_string, ["boolean", "bool", "True", "False"], C_BOOL
+    )
+    candidates += simple_candidates(
+        "float",
+        match_string,
+        [
+            "float",
+            "logarithm",
+            "sine",
+            "cosine",
+            "tangent",
+            "exponential",
+            "complex number",
+            "phase",
+            "ratio of",
+        ],
+        C_FLOAT,
+    )
+
+    candidates += simple_candidates(
+        "str", match_string, ["string", "(sub)string", "sub-string", "substring"], C_STR
+    )
+
+    candidates += simple_candidates("str", match_string, ["name", "names"], C_STR_NAMES)
+    ## "? contains 'None if there is no'  --> Union[Null, xxx]"
+    candidates += simple_candidates(
+        "None",
+        match_string,
+        ["``None``", "None"],
+        C_NONE_RETURN,
+        exclude=["previous value", "if there is no"],
+    )
+
+    candidates += object_candidates(match_string, C_OBJECTS)
+
+    return candidates
+
+
+def return_type_from_context(
+    *, docstring: Union[str, List[str]], signature: str, module: str, literal: bool = False
+):
+    try:
+        return str(
+            _type_from_context(
+                module=module, signature=signature, docstring=docstring, literal=literal
+            )["type"]
+        )
+    except Exception:
+        return "Incomplete"
+
+
+def _type_from_context(
+    *, docstring: Union[str, List[str]], signature: str, module: str, literal: bool = False
+):  # -> Dict[str , Union[str,float]]:
+    """Determine the return type of a function or method based on:
+     - the function signature
+     - the terminology used in the docstring
+
+    Logic:
+    - if the signature contains a return type --> <something> then that is returned
+    - use re to find phrases such as:
+
+        - 'Returns ..... '
+        - 'Gets  ..... '
+
+    - docstring is joined without newlines to simplify parsing
+    - then parses the docstring to find references to known types and give then a rating though a hand coded model ()
+    - builds a list return type candidates
+    - selects the highest ranking candidate
+    - the default Type is 'Any'
+    """
+
+    if isinstance(docstring, list):
+        # join with space to avoid ending at a newline
+        docstring = " ".join(docstring)
+
+    # give the regex that searches for returns a 0.2 boost as that is bound to be more relevant
+
+    weighted_regex = (
+        [
+            (RE_LIT_AS_A, 1.0),
+            (RE_LIT_SENTENCE, 2.0),
+        ]
+        if literal
+        else [
+            (RE_RETURN_VALUE, WEIGHT_RETURN_VAL),
+            (RE_RETURN, WEIGHT_RETURNS),
+            (RE_GETS, WEIGHT_GETS),
+            #       (reads_regex, 1.0),
+        ]
+    )
+    # only the function name without the leading module
+    function_re = re.compile(r"[\w|.]+(?=\()")
+
+    # matches: List[re.Match] = []
+    candidates: List[Dict] = [{"match": "default", "type": "Incomplete", "confidence": 0}]
+
+    # if the signature contains a return type , then use that and do nothing else.
+    if "->" in signature:
+        sig_type = signature.split("->")[-1].strip(": ")
+        return {"type": sig_type, "confidence": WEIGHT_LOOPUPS, "match": signature}
+
+    # ------------------------------------------------------
+    # lookup returns that cannot be found based on the docstring from the lookup list
+    try:
+        function_name = function_re.findall(signature)[0]
+    except IndexError:
+        function_name = signature.strip().strip(":()")
+
+    function_name = ".".join((module, function_name))
+
+    if function_name in LOOKUP_LIST.keys():
+        sig_type = LOOKUP_LIST[function_name][0]
+        return {
+            "type": sig_type,
+            "confidence": C_LOOKUP * WEIGHT_LOOPUPS,
+            "match": function_name,
+        }
+    # ------------------------------------------------------
+    # parse the docstring for simple start verbs,
+    # and add them as a candidate
+    candidates += has_none_verb(docstring)
+
+    # ------------------------------------------------------
+    # parse the docstring for the regexes and weigh the results accordingly
+    for weighted in weighted_regex:
+        match_iter = re.finditer(weighted[0], docstring, re.MULTILINE | re.IGNORECASE)
+        for match in match_iter:
+            # matches.append(match)
+            distilled = distill_return(match.group("return"))
+            for item in distilled:
+                candidate = {
+                    "match": match,
+                    "type": item["type"],
+                    "confidence": item["confidence"] * weighted[1],  # add search boost
+                }
+                candidates.append(candidate)
+    # Sort
+    candidates = sorted(candidates, key=lambda x: x["confidence"], reverse=True)
+    best = candidates[0]  # best candidate
+
+    # ref: https://docs.python.org/3/library/typing.html#typing.Coroutine
+    # Coroutine[YieldType, SendType, ReturnType]
+    # todo: sanity check against actual code .....
+    if "This is a coroutine" in docstring and "Coroutine" not in str(best["type"]):  # type: ignore
+        best["type"] = f"Coroutine[{best['type']}, Any, Any]"
+
+    # return the best candidate, or Any
+    return best  # best candidate