datalad_tabby/io/load.py

@@ -10,6 +10,8 @@ from typing import (
     List,
 )
 
+from charset_normalizer import from_path as cs_from_path
+
 from .load_utils import (
     _assign_context,
     _compact_obj,
@@ -28,6 +30,7 @@ def load_tabby(
     jsonld: bool = True,
     recursive: bool = True,
     cpaths: List | None = None,
+    encoding: str | None = None,
 ) -> Dict | List:
     """Load a tabby (TSV) record as structured (JSON(-LD)) data
 
@@ -48,11 +51,14 @@ def load_tabby(
 
     With the ``jsonld`` flag, a declared or default JSON-LD context is
     loaded and inserted into the record.
+
+    Encoding used when reading tsv files can be specified as ``encoding``.
     """
     ldr = _TabbyLoader(
         jsonld=jsonld,
         recursive=recursive,
         cpaths=cpaths,
+        encoding=encoding,
     )
     return ldr(src=src, single=single)
 
@@ -63,6 +69,7 @@ class _TabbyLoader:
         jsonld: bool = True,
         recursive: bool = True,
         cpaths: List[Path] | None = None,
+        encoding: str | None = None,
     ):
         std_convention_path = Path(__file__).parent / 'conventions'
         if cpaths is None:
@@ -70,6 +77,7 @@ class _TabbyLoader:
         else:
             cpaths.append(std_convention_path)
         self._cpaths = cpaths
+        self._encoding = encoding
         self._jsonld = jsonld
         self._recursive = recursive
 
@@ -94,8 +102,24 @@ class _TabbyLoader:
                 src=src,
                 trace=trace,
             )
+        if self._encoding is not None:
+            tsv_obj = self._parse_tsv_single(src, encoding=self._encoding)
+        else:
+            try:
+                tsv_obj = self._parse_tsv_single(src)
+            except UnicodeDecodeError:
+                # by default Path.open() uses locale.getencoding()
+                # that didn't work, try guessing
+                encoding = cs_from_path(src).best().encoding
+                tsv_obj = self._parse_tsv_single(src, encoding=encoding)
 
-        with src.open(newline='') as tsvfile:
+        obj.update(tsv_obj)
+
+        return self._postproc_obj(obj, src=src, trace=trace)
+
+    def _parse_tsv_single(self, src: Path, encoding: str | None = None) -> Dict:
+        obj = {}
+        with src.open(newline='', encoding=encoding) as tsvfile:
             reader = csv.reader(tsvfile, delimiter='\t')
             # row_id is useful for error reporting
             for row_id, row in enumerate(reader):
@@ -117,8 +141,7 @@ class _TabbyLoader:
                 # we support "sequence" values via multi-column values
                 # supporting two ways just adds unnecessary complexity
                 obj[key] = val
-
+        return obj
-        return self._postproc_obj(obj, src=src, trace=trace)
 
     def _load_many(
         self,
@@ -144,19 +167,49 @@ class _TabbyLoader:
 
         # the table field/column names have purposefully _nothing_
         # to do with any possibly loaded JSON data
-        fieldnames = None
+        if self._encoding is not None:
+            tsv_array = self._parse_tsv_many(
+                src, obj_tmpl, trace=trace, fieldnames=None, encoding=self._encoding
+            )
+        else:
+            try:
+                tsv_array = self._parse_tsv_many(
+                    src, obj_tmpl, trace=trace, fieldnames=None
+                )
+            except UnicodeDecodeError:
+                # by default Path.open() uses locale.getencoding()
+                # that didn't work, try guessing
+                encoding = cs_from_path(src).best().encoding
+                tsv_array = self._parse_tsv_many(
+                    src, obj_tmpl, trace=trace, fieldnames=None, encoding=encoding
+                )
 
-        with src.open(newline='') as tsvfile:
+        array.extend(tsv_array)
+
+        return array
+
+    def _parse_tsv_many(
+        self,
+        src: Path,
+        obj_tmpl: Dict,
+        trace: List,
+        fieldnames: List | None = None,
+        encoding: str | None = None,
+    ) -> List[Dict]:
+        array = []
+        with src.open(newline="", encoding=encoding) as tsvfile:
             # we cannot use DictReader -- we need to support identically named
             # columns
-            reader = csv.reader(tsvfile, delimiter='\t')
+            reader = csv.reader(tsvfile, delimiter="\t")
             # row_id is useful for error reporting
             for row_id, row in enumerate(reader):
                 # row is a list of field, with only as many items
                 # as this particular row has columns
-                if not len(row) \
+                if (
-                        or row[0].startswith('#') \
+                    not len(row)
-                        or all(v is None for v in row):
+                    or row[0].startswith("#")
+                    or all(v is None for v in row)
+                ):
                     # skip empty rows, rows with no key, or rows with
                     # a comment key
                     continue

setup.cfg

@@ -17,6 +17,7 @@ install_requires =
     datalad >= 0.18.0
     datalad-next @ git+https://github.com/datalad/datalad-next.git@main
     datalad-metalad
+    charset-normalizer
     openpyxl
     pyld
 packages = find_namespace: