sonatoki 0.6.2__tar.gz → 0.7.0__tar.gz

{sonatoki-0.6.2 → sonatoki-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sonatoki
-Version: 0.6.2
+Version: 0.7.0
 Summary: ilo li moku e toki li pana e sona ni: ni li toki ala toki pona?
 Author-Email: "jan Kekan San (@gregdan3)" <gregory.danielson3@gmail.com>
 License: AGPL-3.0-or-later

{sonatoki-0.6.2 → sonatoki-0.7.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "sonatoki"
-version = "0.6.2"
+version = "0.7.0"
 description = "ilo li moku e toki li pana e sona ni: ni li toki ala toki pona?"
 authors = [
     { name = "jan Kekan San (@gregdan3)", email = "gregory.danielson3@gmail.com" },

{sonatoki-0.6.2 → sonatoki-0.7.0}/src/sonatoki/Configs.py

@@ -1,49 +1,74 @@
 # STL
-from copy import deepcopy
-from typing import Set, List, Type, TypedDict, cast
+from typing import List, Type, TypedDict
 
 # PDM
 from typing_extensions import NotRequired
 
 # LOCAL
+from sonatoki.types import Number
 from sonatoki.Filters import (
     Or,
     And,
     Not,
     Filter,
     Numeric,
-    Syllabic,
     NimiUCSUR,
     Alphabetic,
     NimiKuLili,
     NimiKuSuli,
     ProperName,
-    Phonotactic,
     Punctuation,
     LongSyllabic,
     Miscellaneous,
-    NimiLinkuCore,
     LongAlphabetic,
     LongProperName,
-    NimiLinkuCommon,
     FalsePosSyllabic,
+    NimiLinkuByUsage,
     NimiLinkuObscure,
     NimiLinkuSandbox,
     NimiLinkuUncommon,
     FalsePosAlphabetic,
 )
-from sonatoki.Scorers import Number, Scorer, PassFail, SoftScaling, SoftPassFail
+from sonatoki.Scorers import Scorer, PassFail, SoftScaling, SoftPassFail
 from sonatoki.Cleaners import Cleaner, ConsecutiveDuplicates
 from sonatoki.Tokenizers import Tokenizer, WordTokenizerRe
 from sonatoki.Preprocessors import (
     URLs,
     Emoji,
-    Backticks,
+    Codeblock,
     Reference,
     Preprocessor,
     AngleBracketObject,
 )
 
+__DICT_PHONOMATCHES = {
+    # Sandbox words are removed from the CorpusConfig if they appear more frequently in English than Toki Pona by a factor of at least 3.
+    # In this case, all of these appear more often in English by a factor of at least 10.
+    "aka",  # also known as
+    "an",  # article
+    "api",  # API
+    "i",  # 1st person
+    "kana",  # japanese script
+    "me",  # 1st person singular, english
+    "ne",  # "no" in several languages
+    "nu",  # "new" in english, "now" in dutch
+    "se",  # spanish particle, english "see"
+    "take",  # acquire, perhaps forcefully or without permission
+    "ten",  # 10
+    "to",  # to, too
+    "je",  # 1st person pronoun, french
+    "u",  # no u
+    "we",  # 1st person plural, english
+    "wi",  # wii and discussions of syllables
+    "sole",  # singular, of shoe
+    # unexplored candidates for removal
+    # "omen",  # ominous
+    # "papa",  # father
+    # "lo",  # "lo" and "loo"
+    # "ewe",  # sheep
+    # "pa",  # father- eh?
+}
+
 
 class IloConfig(TypedDict):
     preprocessors: List[Type[Preprocessor]]
@@ -69,11 +94,11 @@ BaseConfig: IloConfig = {
 
 
 PrefConfig: IloConfig = {
-    "preprocessors": [Emoji, Backticks, URLs, AngleBracketObject, Reference],
+    "preprocessors": [Emoji, Codeblock, URLs, AngleBracketObject, Reference],
     "cleaners": [ConsecutiveDuplicates],
     "ignoring_filters": [Numeric, Punctuation],
     "scoring_filters": [
-        Or(NimiLinkuCore, NimiLinkuCommon, NimiLinkuUncommon, NimiUCSUR),
+        Or(NimiLinkuByUsage(30), NimiUCSUR),
         And(LongSyllabic, Not(FalsePosSyllabic)),
         # NOTE: These are allowed to pass name and alphabetic below, because they *could* be wrong
         LongProperName,
@@ -84,16 +109,13 @@ PrefConfig: IloConfig = {
 }
 
 CorpusConfig: IloConfig = {
-    "preprocessors": [Emoji, Backticks, URLs, AngleBracketObject, Reference],
+    "preprocessors": [Emoji, Codeblock, URLs, AngleBracketObject, Reference],
     "cleaners": [ConsecutiveDuplicates],
     "ignoring_filters": [Numeric, Punctuation],
     "scoring_filters": [
         Or(
-            NimiLinkuCore,
-            NimiLinkuCommon,
-            NimiLinkuUncommon,
-            NimiLinkuObscure,
-            NimiLinkuSandbox,
+            # awkward but efficient syntax
+            NimiLinkuByUsage(0)(sub=__DICT_PHONOMATCHES),
             NimiUCSUR,
             Miscellaneous,
         ),
@@ -104,43 +126,9 @@ CorpusConfig: IloConfig = {
     "scorer": SoftScaling,
     "passing_score": 0.8,
 }
-
-# TODO: create a mechanism to omit tokens from a filter with more granularity
-__corpus_tokens_dict: Set[str] = cast(
-    Set[str],
-    CorpusConfig["scoring_filters"][
-        0
-    ].tokens,  # pyright: ignore[reportAttributeAccessIssue]
-)
-__corpus_tokens_dict -= {
-    # Sandbox words are removed from the CorpusConfig if they appear more frequently in English than Toki Pona by a factor of at least 3.
-    # In this case, all of these appear more often in English by a factor of at least 10.
-    "aka",  # also known as
-    "an",  # article
-    "api",  # API
-    "i",  # 1st person
-    "kana",  # japanese script
-    "me",  # 1st person
-    "ne",  # "no" in several languages
-    "nu",  # "new", now in dutch
-    "se",  # spanish particle, "see"
-    "take",  # acquire, perhaps forcefully or without permission
-    "ten",  # 10
-    "to",  # to, too
-    "u",  # no u
-    "we",  # 1st person plural
-    "wi",  # wii and discussions of syllables
-    "sole",  # singular, of shoe
-    # unexplored candidates for removal
-    # "omen",  # ominous
-    # "papa",  # father
-    # "lo",  # "lo" and "loo"
-    # "ewe",  # sheep
-    # "pa",  # father- eh?
-}
 """Mimics the previous implementation of ilo pi toki pona taso."""
 LazyConfig: IloConfig = {
-    "preprocessors": [Emoji, Backticks, URLs, AngleBracketObject, Reference],
+    "preprocessors": [Emoji, Codeblock, URLs, AngleBracketObject, Reference],
     "cleaners": [ConsecutiveDuplicates],
     "ignoring_filters": [Numeric, Punctuation],
     "scoring_filters": [Alphabetic, NimiUCSUR, ProperName, Miscellaneous],
@@ -150,7 +138,7 @@ LazyConfig: IloConfig = {
 }
 """This is extremely silly."""
 IsipinEpikuConfig: IloConfig = {
-    "preprocessors": [Emoji, Backticks, URLs, AngleBracketObject, Reference],
+    "preprocessors": [Emoji, Codeblock, URLs, AngleBracketObject, Reference],
     "cleaners": [ConsecutiveDuplicates],
     "ignoring_filters": [Numeric, Punctuation],
     "scoring_filters": [
@@ -170,31 +158,10 @@ IsipinEpikuConfig: IloConfig = {
 }
 
 
-DiscordConfig: IloConfig = {
-    "preprocessors": [Emoji, Backticks, URLs, AngleBracketObject, Reference],
-    "cleaners": [ConsecutiveDuplicates],
-    "ignoring_filters": [Numeric, Punctuation],
-    "scoring_filters": [
-        Or(NimiLinkuCore, NimiLinkuCommon, NimiUCSUR, Miscellaneous),
-        And(LongSyllabic, Not(FalsePosSyllabic)),
-        LongProperName,
-        And(LongAlphabetic, Not(FalsePosAlphabetic)),
-    ],
-    "scorer": SoftScaling,
-    "passing_score": 0.8,
-}
-
-TelegramConfig: IloConfig = deepcopy(PrefConfig)
-ForumConfig: IloConfig = deepcopy(PrefConfig)
-
-
 __all__ = [
     "BaseConfig",
     "CorpusConfig",
-    "DiscordConfig",
-    "ForumConfig",
     "IloConfig",
     "LazyConfig",
     "PrefConfig",
-    "TelegramConfig",
 ]

{sonatoki-0.6.2 → sonatoki-0.7.0}/src/sonatoki/Filters.py

@@ -1,37 +1,33 @@
 # STL
 import re
 from abc import ABC, abstractmethod
-from typing import Set, List, Type
+from copy import deepcopy
+from typing import Set, List, Type, Union, Literal, Optional
 from functools import lru_cache as cache  # cache comes in 3.9
 
 # PDM
 import regex
-from typing_extensions import override, deprecated
+from typing_extensions import override
 
 # LOCAL
+from sonatoki.types import LinkuBooks, LinkuUsageDate, LinkuUsageCategory
 from sonatoki.utils import prep_dictionary
 from sonatoki.constants import (
     VOWELS,
-    NIMI_PU,
     ALPHABET,
     ALL_PUNCT,
     ALLOWABLES,
     CONSONANTS,
     NIMI_UCSUR,
-    NIMI_KU_LILI,
-    NIMI_KU_SULI,
-    NIMI_LINKU_CORE,
     NIMI_PU_SYNONYMS,
-    NIMI_LINKU_COMMON,
     FALSE_POS_SYLLABIC,
-    NIMI_LINKU_OBSCURE,
-    NIMI_LINKU_SANDBOX,
     NOT_IN_PUNCT_CLASS,
-    NIMI_LINKU_UNCOMMON,
     ALL_PUNCT_RANGES_STR,
     FALSE_POS_ALPHABETIC,
     UCSUR_PUNCT_RANGES_STR,
     EMOJI_VARIATION_SELECTOR_RANGES_STR,
+    words_by_tag,
+    words_by_usage,
 )
 
 regex.DEFAULT_VERSION = regex.VERSION1
@@ -101,6 +97,20 @@ class MemberFilter(Filter):
     def filter(cls, token: str) -> bool:
         return token.lower() in cls.tokens
 
+    def __new__(
+        cls, add: Optional[Set[str]] = None, sub: Optional[Set[str]] = None
+    ) -> Type[Filter]:
+        parent_tokens = deepcopy(cls.tokens)
+        if add:
+            parent_tokens = parent_tokens.union(add)
+        if sub:
+            parent_tokens -= sub
+
+        class AnonMemberFilter(MemberFilter):
+            tokens = parent_tokens
+
+        return AnonMemberFilter
+
 
 class SubsetFilter(Filter):
     tokens: Set[str]
@@ -155,40 +165,46 @@ class LongProperName(MinLen, ProperName):
     length = 2  # reject "names" of length 1
 
 
-class NimiPu(MemberFilter):
-    tokens = prep_dictionary(NIMI_PU)
-
-
-class NimiPuSynonyms(MemberFilter):
-    tokens = prep_dictionary(NIMI_PU_SYNONYMS)
+class NimiLinkuByUsage:
+    def __new__(
+        cls,
+        usage: int,
+        date: Optional[LinkuUsageDate] = None,
+    ) -> Type[MemberFilter]:
+        words = words_by_usage(usage, date)
 
+        class AnonLinkuMemberFilter(MemberFilter):
+            tokens = prep_dictionary(words)
 
-class NimiKuSuli(MemberFilter):
-    tokens = prep_dictionary(NIMI_KU_SULI)
+        return AnonLinkuMemberFilter
 
 
-class NimiKuLili(MemberFilter):
-    tokens = prep_dictionary(NIMI_KU_LILI)
+class NimiLinkuByTag:
+    def __new__(
+        cls,
+        tag: Union[Literal["usage_category"], Literal["book"]],
+        category: Union[LinkuUsageCategory, LinkuBooks],
+    ) -> Type[MemberFilter]:
+        words = words_by_tag(tag, category)
 
+        class AnonLinkuMemberFilter(MemberFilter):
+            tokens = prep_dictionary(words)
 
-class NimiLinkuCore(MemberFilter):
-    tokens = prep_dictionary(NIMI_LINKU_CORE)
+        return AnonLinkuMemberFilter
 
 
-class NimiLinkuCommon(MemberFilter):
-    tokens = prep_dictionary(NIMI_LINKU_COMMON)
+NimiPu = NimiLinkuByTag("book", "pu")
+NimiKuSuli = NimiLinkuByTag("book", "ku suli")
+NimiKuLili = NimiLinkuByTag("book", "ku lili")
+NimiLinkuCore = NimiLinkuByTag("usage_category", "core")
+NimiLinkuCommon = NimiLinkuByTag("usage_category", "common")
+NimiLinkuUncommon = NimiLinkuByTag("usage_category", "uncommon")
+NimiLinkuObscure = NimiLinkuByTag("usage_category", "obscure")
+NimiLinkuSandbox = NimiLinkuByTag("usage_category", "sandbox")
 
 
-class NimiLinkuUncommon(MemberFilter):
-    tokens = prep_dictionary(NIMI_LINKU_UNCOMMON)
-
-
-class NimiLinkuObscure(MemberFilter):
-    tokens = prep_dictionary(NIMI_LINKU_OBSCURE)
-
-
-class NimiLinkuSandbox(MemberFilter):
-    tokens = prep_dictionary(NIMI_LINKU_SANDBOX)
+class NimiPuSynonyms(MemberFilter):
+    tokens = prep_dictionary(NIMI_PU_SYNONYMS)
 
 
 class NimiUCSUR(MemberFilter):

{sonatoki-0.6.2 → sonatoki-0.7.0}/src/sonatoki/Preprocessors.py

@@ -143,6 +143,15 @@ class Backticks(RegexPreprocessor):
     pattern = re.compile(r"`[^`]+`", flags=re.DOTALL)
 
 
+class Codeblock(RegexPreprocessor):
+    """Remove codeblocks marked by a set of three backticks on their own lines.
+
+    Subset of what would be removed by Backticks, but may be preferable.
+    """
+
+    pattern = re.compile(r"```\n(?:(?!```).*?)?```", flags=re.DOTALL)
+
+
 class Spoilers(RegexPreprocessor):
     """Remove paired double bars and their contents `||like this||`"""
 

{sonatoki-0.6.2 → sonatoki-0.7.0}/src/sonatoki/Scorers.py

@@ -1,17 +1,15 @@
 # STL
 import math
 from abc import ABC, abstractmethod
-from typing import Dict, List, Type, Union
+from typing import List, Type
 
 # PDM
 from typing_extensions import override
 
 # LOCAL
+from sonatoki.types import Number, Scorecard
 from sonatoki.Filters import Filter
 
-Number = Union[int, float]
-Weights = Dict[str, Number]
-
 
 class Scorer(ABC):
     @classmethod
@@ -124,7 +122,64 @@ class SoftScaling(Soften, Scaling):
     scoring."""
 
 
-# class Logarithmic(Scorer): ...
+class SentenceScorer(ABC):
+    @classmethod
+    @abstractmethod
+    def score(cls, scorecards: List[Scorecard]) -> List[Scorecard]:
+        """Re-score a list of sentences (scorecards, sentences with all their
+        metadata) and return them."""
+        raise NotImplementedError
+
+
+class SentNoOp(SentenceScorer):
+    @classmethod
+    @override
+    def score(cls, scorecards: List[Scorecard]) -> List[Scorecard]:
+        return scorecards
 
 
-__all__ = ["PassFail", "SoftPassFail", "Scaling", "SoftScaling"]
+class SentAvg(SentenceScorer):
+    @classmethod
+    @override
+    def score(cls, scorecards: List[Scorecard]) -> List[Scorecard]:
+        if not scorecards:
+            return scorecards
+
+        total = sum(card["score"] for card in scorecards)
+        avg = total / len(scorecards)
+        for card in scorecards:
+            card["score"] = avg
+        return scorecards
+
+
+class SentWeightedAvg(SentenceScorer):
+    @classmethod
+    @override
+    def score(cls, scorecards: List[Scorecard]) -> List[Scorecard]:
+        if not scorecards:
+            return scorecards
+
+        weighted_total = 0
+        total_len = 0
+        for card in scorecards:
+            cardlen = len(card["cleaned"])
+            cardscore = card["score"]
+
+            weighted_total += cardlen * cardscore
+            total_len += cardlen
+
+        weighted_avg = weighted_total / total_len
+        for card in scorecards:
+            card["score"] = weighted_avg
+        return scorecards
+
+
+__all__ = [
+    "PassFail",
+    "Scaling",
+    "SoftPassFail",
+    "SoftScaling",
+    "Soften",
+    "SentAvg",
+    "SentWeightedAvg",
+]

{sonatoki-0.6.2 → sonatoki-0.7.0}/src/sonatoki/constants.py

@@ -1,11 +1,16 @@
 # STL
 import json
-from typing import Set, Dict
+from typing import Set, Dict, Optional
 from pathlib import Path
 
 # LOCAL
+from sonatoki.types import LinkuWord, LinkuUsageDate
 from sonatoki.utils import find_unicode_chars, find_unicode_ranges
 
+LATEST_DATE = "2023-09"
+# hardcoding this seems bad, but it means the parser is stable w.r.t. Linku!
+
+
 # `\p{Punctuation}` character class
 # https://www.compart.com/en/unicode/category
 # https://unicode.org/Public/UNIDATA/UnicodeData.txt
@@ -638,6 +643,7 @@ FALSE_POS_SYLLABIC = {
     "iluminate",
     "imense",
     "imitate",
+    "inanimate",
     "injoke",
     "insane",
     "insolate",
@@ -689,26 +695,42 @@ NIMI_UCSUR = find_unicode_chars(UCSUR_RANGES)
 # NIMI_PU_ALE_UCSUR_RANGES = NIMI_PU_UCSUR_RANGES + ["\\U000F1978-\\U000F197A"]
 
 
-def category_helper(data: Dict[str, Dict[str, str]], key: str, value: str) -> Set[str]:
-    return {d["word"] for d in data.values() if d[key] == value}
+def linku_data() -> Dict[str, LinkuWord]:
+    # NOTE: this does open+read+parse two files each time you construct a filter
+    # but i expect users to construct filters only at the start of runtime
+    # there is no reason to waste your RAM by leaving the linku data in it
+    with open(LINKU) as f:
+        linku: Dict[str, LinkuWord] = json.loads(f.read())
+    with open(SANDBOX) as f:
+        sandbox: Dict[str, LinkuWord] = json.loads(f.read())
+
+    return {**linku, **sandbox}
+
 
+def words_by_tag(tag: str, value: str) -> Set[str]:
+    data = linku_data()
+    return {d["word"] for d in data.values() if d[tag] == value}
 
-with open(LINKU) as f:
-    linku: Dict[str, Dict[str, str]] = json.loads(f.read())
-    NIMI_PU = category_helper(linku, "book", "pu")
-    NIMI_PU_SYNONYMS = {"namako", "kin", "oko"}
 
-    NIMI_KU_SULI = category_helper(linku, "book", "ku suli")
-    NIMI_KU_LILI = category_helper(linku, "book", "ku lili")
+def words_by_usage(
+    usage: int,
+    date: Optional[LinkuUsageDate] = None,
+) -> Set[str]:
+    if not date:
+        date = LATEST_DATE
+    data = linku_data()
 
-    NIMI_LINKU_CORE = category_helper(linku, "usage_category", "core")
-    NIMI_LINKU_COMMON = category_helper(linku, "usage_category", "common")
-    NIMI_LINKU_UNCOMMON = category_helper(linku, "usage_category", "uncommon")
-    NIMI_LINKU_OBSCURE = category_helper(linku, "usage_category", "obscure")
+    result: Set[str] = set()
+    for word in data.values():
+        usages = word["usage"]
+        if date in usages and usages[date] >= usage:
+            result.add(word["word"])
+
+    return result
+
+
+NIMI_PU_SYNONYMS = {"namako", "kin", "oko"}
 
-with open(SANDBOX) as f:
-    sandbox: Dict[str, Dict[str, str]] = json.loads(f.read())
-    NIMI_LINKU_SANDBOX = {d["word"] for d in sandbox.values()}
 
 # with open(SYLLABICS) as f:
 #     FALSE_POS_SYLLABIC = {line.strip() for line in f}
@@ -716,9 +738,6 @@ with open(SANDBOX) as f:
 # with open(ALPHABETICS) as f:
 #     FALSE_POS_ALPHABETIC = {line.strip() for line in f}
 
-del linku
-del sandbox
-
 __all__ = [
     "ALLOWABLES",
     "ALL_PUNCT",
@@ -727,14 +746,6 @@ __all__ = [
     "CONSONANTS",
     "EMOJI_VARIATION_SELECTOR_RANGES",
     "EMOJI_VARIATION_SELECTOR_RANGES_STR",
-    "NIMI_KU_LILI",
-    "NIMI_KU_SULI",
-    "NIMI_LINKU_COMMON",
-    "NIMI_LINKU_CORE",
-    "NIMI_LINKU_OBSCURE",
-    "NIMI_LINKU_SANDBOX",
-    "NIMI_LINKU_UNCOMMON",
-    "NIMI_PU",
     "NIMI_PU_SYNONYMS",
     "POSIX_PUNCT",
     "POSIX_PUNCT_RANGES",

{sonatoki-0.6.2 → sonatoki-0.7.0}/src/sonatoki/ilo.py

@@ -1,17 +1,14 @@
 # STL
-from typing import List, Type, Tuple
+from typing import List, Type
 
 # LOCAL
+from sonatoki.types import Number, Scorecard
 from sonatoki.Filters import Filter
-from sonatoki.Scorers import Number, Scorer
+from sonatoki.Scorers import Scorer, SentNoOp, SentenceScorer
 from sonatoki.Cleaners import Cleaner
 from sonatoki.Tokenizers import Tokenizer, SentTokenizer, WordTokenizer
 from sonatoki.Preprocessors import Preprocessor
 
-# tokenized, filtered, cleaned, score, result
-Scorecard = Tuple[List[str], List[str], List[str], Number, bool]
-# TODO: scorecard kinda sucks as a name
-
 
 class Ilo:
     __preprocessors: List[Type[Preprocessor]]
@@ -21,6 +18,7 @@ class Ilo:
     __ignoring_filters: List[Type[Filter]]
     __scoring_filters: List[Type[Filter]]
     __scorer: Type[Scorer]
+    __sentence_scorer: Type[SentenceScorer]
     __passing_score: Number
 
     def __init__(
@@ -31,6 +29,7 @@ class Ilo:
         scoring_filters: List[Type[Filter]],
         scorer: Type[Scorer],
         passing_score: Number,
+        sentence_scorer: Type[SentenceScorer] = SentNoOp,
         word_tokenizer: Type[Tokenizer] = WordTokenizer,
         sent_tokenizer: Type[Tokenizer] = SentTokenizer,
     ):
@@ -43,6 +42,7 @@ class Ilo:
         self.__ignoring_filters = [*ignoring_filters]
         self.__scoring_filters = [*scoring_filters]
         self.__scorer = scorer
+        self.__sentence_scorer = sentence_scorer
         self.__passing_score = passing_score
 
     def preprocess(self, msg: str) -> str:
@@ -55,6 +55,7 @@ class Ilo:
         return self.__word_tokenizer.tokenize(msg)
 
     def sent_tokenize(self, msg: str) -> List[str]:
+        """It is *highly* recommended that you run `ilo.preprocess` first."""
         return self.__sent_tokenizer.tokenize(msg)
 
     def clean_token(self, token: str) -> str:
@@ -93,44 +94,50 @@ class Ilo:
     def score_tokens(self, tokens: List[str]) -> float:
         return self.__scorer.score(tokens, self.__scoring_filters)
 
+    def score_sentences(self, scorecards: List[Scorecard]) -> List[Scorecard]:
+        return self.__sentence_scorer.score(scorecards)
+
     def _is_toki_pona(self, message: str) -> Scorecard:
         """Process a message into its tokens, then filters, cleans, and scores
-        them. Returns all parts. Message must already be preprocessed, normally
-        done in `self.is_toki_pona(message)`.
-
-        Returns all components of the processing algorithm except preprocessing:
-        - Tokenized message (list[str])
-        - Filtered message (list[str])
-        - Cleaned message (list[str])
-        - Score (float)
-        - Result (bool)
+        them. Message must already be preprocessed, normally done in
+        `self.is_toki_pona(message)`.
+
+        Returns a `Scorecard` with all changes to the input text and a score.
         """
         tokenized = self.word_tokenize(message)
         filtered = self.filter_tokens(tokenized)
         cleaned = self.clean_tokens(filtered)
         score = self.score_tokens(cleaned)
-        result = score >= self.__passing_score
 
-        return tokenized, filtered, cleaned, score, result
+        scorecard: Scorecard = {
+            "text": message,
+            "tokenized": tokenized,
+            "filtered": filtered,
+            "cleaned": cleaned,
+            "score": score,
+        }
+
+        return scorecard
 
     def is_toki_pona(self, message: str) -> bool:
-        """Determines whether a single statement is or is not Toki Pona."""
+        """Determines whether a text is or is not Toki Pona."""
         message = self.preprocess(message)
-        *_, result = self._is_toki_pona(message)
-        return result
+        scorecard = self._is_toki_pona(message)
+        return scorecard["score"] >= self.__passing_score
 
     def _are_toki_pona(self, message: str) -> List[Scorecard]:
-        """Split a message into sentences, then return a list each sentence's
-        results via `self._is_toki_pona()`.
+        """Split a message into sentences, then return a list with each
+        sentence's scorecard from `self._is_toki_pona()`.
 
         Message must already be preprocessed, normally done in
         `self.are_toki_pona(message)`.
         """
-        results: List[Scorecard] = list()
+        scorecards: List[Scorecard] = list()
         for sentence in self.sent_tokenize(message):
             result = self._is_toki_pona(sentence)
-            results.append(result)
-        return results
+            scorecards.append(result)
+        scorecards = self.score_sentences(scorecards)
+        return scorecards
 
     def are_toki_pona(self, message: str) -> List[bool]:
         """Splits a statement into sentences, then determines if each is or is not Toki Pona.
@@ -148,5 +155,5 @@ class Ilo:
         ```
         """
         message = self.preprocess(message)
-        results = self._are_toki_pona(message)
-        return [res[-1] for res in results]
+        scorecards = self._are_toki_pona(message)
+        return [card["score"] >= self.__passing_score for card in scorecards]

sonatoki-0.7.0/src/sonatoki/types.py

@@ -0,0 +1,60 @@
+# STL
+from typing import Dict, List, Union, Literal, TypedDict
+
+Number = Union[int, float]
+
+
+# TODO: scorecard kinda sucks as a name
+class Scorecard(TypedDict):
+    text: str
+    tokenized: List[str]
+    filtered: List[str]
+    cleaned: List[str]
+    score: Number
+
+
+LinkuUsageDate = Union[
+    Literal["2020-04"],
+    Literal["2021-10"],
+    Literal["2022-08"],
+    Literal["2023-09"],
+    # Literal["2024-09"],
+]
+
+LinkuUsageCategory = Union[
+    Literal["core"],
+    Literal["common"],
+    Literal["uncommon"],
+    Literal["obscure"],
+    Literal["sandbox"],
+]
+
+LinkuBooks = Union[
+    Literal["pu"],
+    Literal["ku suli"],
+    Literal["ku lili"],
+    Literal["none"],
+]
+
+
+class LinkuWord(TypedDict):
+    id: str
+    author_verbatim: str
+    author_verbatim_source: str
+    book: str
+    coined_era: str
+    coined_year: str
+    creator: List[str]
+    ku_data: Dict[str, int]
+    see_also: List[str]
+    resources: Dict[str, str]
+    representations: Dict[str, Union[str, List[str]]]
+    source_language: str
+    usage_category: LinkuUsageCategory
+    word: str
+    deprecated: bool
+    etymology: List[Dict[str, str]]
+    audio: List[Dict[str, str]]
+    pu_verbatim: Dict[str, str]
+    usage: Dict[LinkuUsageDate, int]
+    translations: Dict[str, Dict[str, str]]

{sonatoki-0.6.2 → sonatoki-0.7.0}/tests/test_filters.py

@@ -34,23 +34,13 @@ from sonatoki.Filters import (
     NimiLinkuUncommon,
 )
 from sonatoki.Cleaners import Lowercase, ConsecutiveDuplicates
-from sonatoki.constants import (
-    NIMI_PU,
-    NIMI_KU_LILI,
-    NIMI_KU_SULI,
-    NIMI_LINKU_CORE,
-    NIMI_LINKU_COMMON,
-    FALSE_POS_SYLLABIC,
-    NIMI_LINKU_OBSCURE,
-    NIMI_LINKU_SANDBOX,
-    NIMI_LINKU_UNCOMMON,
-)
+from sonatoki.constants import FALSE_POS_SYLLABIC, words_by_tag
 
 # FILESYSTEM
 from .test_utils import PROPER_NAME_RE
 
 
-@given(st.sampled_from(list(NIMI_PU)))
+@given(st.sampled_from(list(words_by_tag("book", "pu"))))
 @example("lukin")
 @example("selo")
 @example("li")
@@ -59,14 +49,14 @@ def test_NimiPu(s: str):
     assert res, repr(s)
 
 
-@given(st.sampled_from(list(NIMI_LINKU_CORE)))
+@given(st.sampled_from(list(words_by_tag("usage_category", "core"))))
 @example("pona")
 def test_NimiLinkuCore(s: str):
     res = NimiLinkuCore.filter(s)
     assert res, repr(s)
 
 
-@given(st.sampled_from(list(NIMI_LINKU_COMMON)))
+@given(st.sampled_from(list(words_by_tag("usage_category", "common"))))
 @example("n")
 @example("tonsi")
 @example("kipisi")
@@ -75,19 +65,21 @@ def test_NimiLinkuCommon(s: str):
     assert res, repr(s)
 
 
-@given(st.sampled_from(list(NIMI_LINKU_UNCOMMON)))
+@given(st.sampled_from(list(words_by_tag("usage_category", "uncommon"))))
 def test_NimiLinkuUncommon(s: str):
     res = NimiLinkuUncommon.filter(s)
     assert res, repr(s)
 
 
-@given(st.sampled_from(list(NIMI_LINKU_OBSCURE)))
+@given(st.sampled_from(list(words_by_tag("usage_category", "obscure"))))
+@example("pake")
+@example("san")
 def test_NimiLinkuObscure(s: str):
     res = NimiLinkuObscure.filter(s)
     assert res, repr(s)
 
 
-@given(st.sampled_from(list(NIMI_LINKU_SANDBOX)))
+@given(st.sampled_from(list(words_by_tag("usage_category", "sandbox"))))
 @example("kalamARR")
 @example("Pingo")
 def test_NimiLinkuSandbox(s: str):
@@ -207,7 +199,11 @@ def test_OrFilter(s: str):
 # NOTE: No subset filter test because A | B is not the same as A combined with B.
 # e.g. "apple" passes Alphabetic, "..." passes Punctuation, "apple..." passes neither
 # but would incorrectly pass a combined filter.
-@given(st.sampled_from(list(NIMI_PU | NIMI_LINKU_OBSCURE)))
+@given(
+    st.sampled_from(
+        list(words_by_tag("book", "pu") | words_by_tag("usage_category", "obscure"))
+    )
+)
 def test_MemberFilters_OrFilter(s: str):
     filter = Or(NimiPu, NimiLinkuObscure)
     assert issubclass(filter, MemberFilter)
@@ -221,11 +217,11 @@ def test_MemberFilters_OrFilter(s: str):
 @given(
     st.sampled_from(
         list(
-            NIMI_KU_SULI
-            | NIMI_KU_LILI
-            | NIMI_LINKU_UNCOMMON
-            | NIMI_LINKU_OBSCURE
-            | NIMI_LINKU_SANDBOX
+            words_by_tag("book", "ku suli")
+            | words_by_tag("book", "ku lili")
+            | words_by_tag("usage_category", "uncommon")
+            | words_by_tag("usage_category", "obscure")
+            | words_by_tag("usage_category", "sandbox")
         ),
     )
 )
@@ -248,14 +244,14 @@ def test_OrFilter_IsipinEpiku(s: str):
     )
 
 
-@given(st.sampled_from(list(NIMI_PU)))
+@given(st.sampled_from(list(words_by_tag("book", "pu"))))
 def test_AndFilter(s: str):
     s = s.capitalize()
     f = And(ProperName, NimiPu)
     assert f.filter(s)
 
 
-@given(st.sampled_from(list(NIMI_PU)))
+@given(st.sampled_from(list(words_by_tag("book", "pu"))))
 def test_NotFilter(s: str):
     f = Not(NimiPu)
     assert not f.filter(s)
@@ -280,3 +276,44 @@ def test_AndNotFilter(s: str):
     if res_fp:
         # syl matched- but if fp matches, then the composed filter should not match
         assert not res_composed
+
+
+@given(
+    st.sampled_from(list(words_by_tag("book", "pu") | words_by_tag("book", "ku suli")))
+)
+def test_AddTokensToMemberFilter(s: str):
+    PuEnKuSuliFilter = NimiPu(add=NimiKuSuli.tokens)
+    assert PuEnKuSuliFilter.filter(s)
+
+
+@given(
+    st.sampled_from(
+        list(
+            words_by_tag("usage_category", "sandbox") | words_by_tag("book", "ku lili")
+        )
+    )
+)
+def test_AddTokensToMemberFilterNegative(s: str):
+    PuEnKuSuliFilter = NimiPu(add=NimiKuSuli.tokens)
+    assert not PuEnKuSuliFilter.filter(s)
+
+
+@given(
+    st.sampled_from(
+        list(
+            words_by_tag("book", "pu")
+            | words_by_tag("book", "ku suli")
+            | words_by_tag("book", "ku lili")
+            | words_by_tag("usage_category", "uncommon")
+            | words_by_tag("usage_category", "obscure")
+            | words_by_tag("usage_category", "sandbox")
+        ),
+    )
+    | st.from_regex(Syllabic.pattern.pattern, fullmatch=True)
+)
+def test_SubTokensFromMemberFilter(s: str):
+    NimiAlaFilter = NimiLinkuCore(sub=NimiPu.tokens)
+    # core is a strict subset of pu
+    # if kin becomes core, needs to be corrected
+
+    assert not NimiAlaFilter.filter(s)

{sonatoki-0.6.2 → sonatoki-0.7.0}/tests/test_ilo.py

@@ -1,3 +1,6 @@
+# STL
+from typing import List, Tuple
+
 # PDM
 import pytest
 
@@ -35,6 +38,10 @@ ALL_VALID = [
     "󱥄󱥬󱥩󱤴",  # "o toki tawa mi" in UCSUR
     "󱤴󱤧󱤑󱥍󱦗󱤖󱥡󱦘󱤬󱥭‍󱥡󱥚",
     "󱤑󱦐󱥗󱦜󱦈󱦜󱥉󱦜󱦑󱥄󱤤󱤂󱤉󱥆󱤀",
+    "o lukin, 󱤴󱥬󱥩󱤴󱤧wawa",
+    "ni li sona kiwen",
+    "nimi namako li toki e ale",
+    "mi open mute a",  # mostly eng words
 ]
 
 IGNORABLES = [
@@ -55,10 +62,9 @@ IGNORABLES = [
     "❤️",  # heart
     "😊",
     "👨‍👩‍👧‍👧",  # family emoji with zwj
-    # every non-emoji in
+    # every non-emoji in the writables
     "🄀🄁🄂🄃🄄🄅🄆🄇🄈🄉🄊🄋🄌🄍🄎🄏🄐🄑🄒🄓🄔🄕🄖🄗🄘🄙🄚🄛🄜🄝🄞🄟🄠🄡🄢🄣🄤🄥🄦🄧🄨🄩🄪🄫🄬🄭🄮🄯🄰🄱🄲🄳🄴🄵🄶🄷🄸🄹🄺🄻🄼🄽🄾🄿🅀🅁🅂🅃🅄🅅🅆🅇🅈🅉🅊🅋🅌🅍🅎🅏🅐🅑🅒🅓🅔🅕🅖🅗🅘🅙🅚🅛🅜🅝🅞🅟🅠🅡🅢🅣🅤🅥🅦🅧🅨🅩🅪🅫🅬🅭🅮🅯🅲🅳🅴🅵🅶🅷🅸🅹🅺🅻🅼🅽🆀🆁🆂🆃🆄🆅🆆🆇🆈🆉🆊🆋🆌🆍🆏🆐 🆛🆜🆝🆞🆟🆠🆡🆢🆣🆤🆥🆦🆧🆨🆩🆪🆫🆬🆭🇦🇧🇨🇩🇪🇫🇬🇭🇮🇯🇰🇱🇲🇳🇴🇵🇶🇷🇸🇹🇺🇻🇼🇽🇾🇿",
     "🅰️🅱️🅾️🅱️🅰️",  # blood type emojis
-    # "😃⃢👍",  # sincerely, no idea, but it came up
 ]
 
 SYLLABIC_MATCHES = [
@@ -108,7 +114,7 @@ CORPUS_SPECIFIC = [
     "Pingo",
     "we Luke li alente wa",
 ]
-CORPUS_SPECIFIC_XFAIL = []
+CORPUS_SPECIFIC_XFAIL: List[str] = []
 
 
 EXCESSIVE_SYLLABICS = [
@@ -193,10 +199,20 @@ FALSE_NEGATIVES = [
     "mtue",
     "mi nasa B^)",  # emoticon
     "lete li ike x.x",  # this is an emoticon but passes because 'x' is in Filters.Miscellaneous
+    "😃⃢👍",  # sincerely, no idea, but it came up and it should be omitted by emojis but isn't
 ]
 
 FALSE_POSITIVES = [
-    "Knowing a little toki pona",
+    "Knowing a little toki pona",  # name, dict, alphabet, dict, dict- damn, that's hard.
+]
+
+IGNORABLE_PAIRS: List[Tuple[str, str]] = [
+    ("o lukin e ni: https://example.com/", "o lukin e ni:"),
+    ("ni li nasa anu seme <:musiwawa:198591138591>", "ni li nasa anu seme"),
+    ("seme la ni li toki pona ala https://example.com/", "seme la ni li toki pona ala"),
+    ("```\ndef bad():\n    pass\n``` o lukin e ni", "o lukin e ni"),
+    ("mi tawa tomo telo 💦💦", "mi tawa tomo telo"),
+    ("o lukin e lipu ni: [[wp:Canvassing]]", "o lukin e lipu ni:"),
 ]
 
 
@@ -254,3 +270,33 @@ def test_false_negatives_pref(ilo: Ilo, text: str):
 @pytest.mark.parametrize("text", CORPUS_SPECIFIC_XFAIL)
 def test_false_positives_corpus(corpus_ilo: Ilo, text: str):
     assert not corpus_ilo.is_toki_pona(text)
+
+
+@pytest.mark.parametrize("pair", IGNORABLE_PAIRS)
+def test_pref_ignorable_doesnt_change_score(ilo: Ilo, pair: Tuple[str, str]):
+    with_ignorable, without_ignorable = pair
+    with_ignorable = ilo.preprocess(with_ignorable)
+    without_ignorable = ilo.preprocess(without_ignorable)
+    score_with = ilo._is_toki_pona(with_ignorable)["score"]
+    score_without = ilo._is_toki_pona(without_ignorable)["score"]
+    assert score_with == score_without
+
+
+@pytest.mark.parametrize("pair", IGNORABLE_PAIRS)
+def test_lazy_ignorable_doesnt_change_score(lazy_ilo: Ilo, pair: Tuple[str, str]):
+    with_ignorable, without_ignorable = pair
+    with_ignorable = lazy_ilo.preprocess(with_ignorable)
+    without_ignorable = lazy_ilo.preprocess(without_ignorable)
+    score_with = lazy_ilo._is_toki_pona(with_ignorable)["score"]
+    score_without = lazy_ilo._is_toki_pona(without_ignorable)["score"]
+    assert score_with == score_without
+
+
+@pytest.mark.parametrize("pair", IGNORABLE_PAIRS)
+def test_corpus_ignorable_doesnt_change_score(corpus_ilo: Ilo, pair: Tuple[str, str]):
+    with_ignorable, without_ignorable = pair
+    with_ignorable = corpus_ilo.preprocess(with_ignorable)
+    without_ignorable = corpus_ilo.preprocess(without_ignorable)
+    score_with = corpus_ilo._is_toki_pona(with_ignorable)["score"]
+    score_without = corpus_ilo._is_toki_pona(without_ignorable)["score"]
+    assert score_with == score_without

{sonatoki-0.6.2 → sonatoki-0.7.0}/tests/test_preprocessors.py

@@ -8,6 +8,7 @@ from sonatoki.Preprocessors import (
     Spoilers,
     AllQuotes,
     Backticks,
+    Codeblock,
     Reference,
     ArrowQuote,
     ColonEmotes,
@@ -48,6 +49,25 @@ def test_Backticks(s: str):
     assert res == "", (repr(s), repr(res))
 
 
+@given(st.from_regex(Codeblock.pattern.pattern, fullmatch=True))
+@example(
+    """```
+```"""
+)
+@example(
+    """```
+blocky message
+```
+
+```
+second blocky message
+```"""
+)
+def test_Codeblock(s: str):
+    res = Codeblock.process(s).strip()
+    assert res == "", (repr(s), repr(res))
+
+
 @given(st.from_regex(ArrowQuote.pattern.pattern, fullmatch=True))
 @example("> base")
 @example("> newline\n> newline")

{sonatoki-0.6.2 → sonatoki-0.7.0}/tests/test_properties.py

@@ -19,45 +19,35 @@ from sonatoki.Filters import (
 )
 from sonatoki.Cleaners import Lowercase, ConsecutiveDuplicates
 from sonatoki.constants import (
-    NIMI_PU,
-    NIMI_KU_LILI,
-    NIMI_KU_SULI,
-    NIMI_LINKU_CORE,
     NIMI_PU_SYNONYMS,
-    NIMI_LINKU_COMMON,
     FALSE_POS_SYLLABIC,
-    NIMI_LINKU_OBSCURE,
-    NIMI_LINKU_SANDBOX,
-    NIMI_LINKU_UNCOMMON,
     FALSE_POS_ALPHABETIC,
+    words_by_tag,
+    words_by_usage,
 )
 
 
-@given(st.sampled_from(list(NIMI_PU | NIMI_PU_SYNONYMS)))
+@given(st.sampled_from(list(words_by_tag("book", "pu") | NIMI_PU_SYNONYMS)))
 def test_pu_filters_non_overlap(s: str):
     res_pu = NimiPu.filter(s)
     res_synonyms = NimiPuSynonyms.filter(s)
     assert (res_pu + res_synonyms) == 1
 
 
-@given(st.sampled_from(list(NIMI_KU_SULI | NIMI_KU_LILI)))
+@given(
+    st.sampled_from(
+        list(words_by_tag("book", "ku suli") | words_by_tag("book", "ku lili"))
+    )
+)
 def test_ku_filters_non_overlap(s: str):
+    s = Lowercase.clean(s)
+    s = ConsecutiveDuplicates.clean(s)
     res_ku_suli = NimiKuSuli.filter(s)
     res_ku_lili = NimiKuLili.filter(s)
     assert (res_ku_suli + res_ku_lili) == 1
 
 
-@given(
-    st.sampled_from(
-        list(
-            NIMI_LINKU_CORE
-            | NIMI_LINKU_COMMON
-            | NIMI_LINKU_UNCOMMON
-            | NIMI_LINKU_OBSCURE
-            | NIMI_LINKU_SANDBOX
-        )
-    )
-)
+@given(st.sampled_from(list(words_by_usage(0))))
 def test_linku_filters_non_overlap(s: str):
     _ = assume(s != "su")
 
@@ -73,7 +63,7 @@ def test_linku_filters_non_overlap(s: str):
     assert (res_core + res_common + res_uncommon + res_obscure + res_sandbox) == 1
 
 
-@given(st.sampled_from(list(NIMI_LINKU_CORE | NIMI_LINKU_COMMON | NIMI_LINKU_UNCOMMON)))
+@given(st.sampled_from(list(words_by_usage(30))))
 def test_nimi_linku_properties(s: str):
     assert ConsecutiveDuplicates.clean(s) == s, repr(s)
     assert Alphabetic.filter(s), repr(s)

{sonatoki-0.6.2 → sonatoki-0.7.0}/tests/test_utils.py

@@ -1,17 +1,14 @@
-# STL
-import re
-
 # PDM
 import hypothesis.strategies as st
 
 # LOCAL
 from sonatoki.Filters import Syllabic, Phonotactic, AlphabeticRe
-from sonatoki.constants import NIMI_LINKU_CORE, NIMI_LINKU_COMMON
+from sonatoki.constants import words_by_usage
 
 PROPER_NAME_RE = r"[A-Z][a-z]*"
 
 token_strategy = (
-    st.sampled_from(list(NIMI_LINKU_CORE | NIMI_LINKU_COMMON))
+    st.sampled_from(list(words_by_usage(60)))
     | st.from_regex(Phonotactic.pattern.pattern, fullmatch=True)
     | st.from_regex(Syllabic.pattern.pattern, fullmatch=True)
     | st.from_regex(PROPER_NAME_RE, fullmatch=True)