detect encoding only if it's not provided by the user (#4)

geovaex/io.py

@@ -190,15 +190,20 @@ def to_arrow(file, arrow_file, chunksize=2000000, crs=None, **kwargs):
     with pa.OSFile(arrow_file, 'wb') as sink:
         writer = None
         # by default assume utf-8 encoding
-        encoding = 'utf-8'
-        if not os.path.isfile(file):
-            cpg_file = [os.path.join(file, f) for f in os.listdir(file) if f.endswith('.cpg')]
-            if len(cpg_file) == 1:
-                # if there is a cpg file read the encoding out of its first line
-                cpg_file = cpg_file[0]
-                with open(cpg_file) as f:
-                    encoding = f.readline()
-        for table in to_arrow_table(file, chunksize=chunksize, crs=crs, encoding=encoding, **kwargs):
+        if 'encoding' not in kwargs:
+            encoding = 'utf-8'
+            if not os.path.isfile(file):
+                cpg_file = [os.path.join(file, f) for f in os.listdir(file) if f.endswith('.cpg')]
+                if len(cpg_file) == 1:
+                    # if there is a cpg file read the encoding out of its first line
+                    cpg_file = cpg_file[0]
+                    with open(cpg_file) as f:
+                        encoding = f.readline()
+            arrow_generator = to_arrow_table(file, chunksize=chunksize, crs=crs, encoding=encoding, **kwargs)
+        else:
+            arrow_generator = to_arrow_table(file, chunksize=chunksize, crs=crs, **kwargs)
+
+        for table in arrow_generator:
             b = table.to_batches()
             if writer is None:
                 writer = pa.RecordBatchStreamWriter(sink, b[0].schema)