From 83783874458bd9fe79f8e4124062d8dc1a9f8011 Mon Sep 17 00:00:00 2001
From: Brock <jbrockmendel@gmail.com>
Date: Fri, 28 Apr 2023 20:24:53 -0700
Subject: [PATCH] REF: define _header_line and have_mi_columns non-dynamically

---
 pandas/io/parsers/python_parser.py | 54 ++++++++++++++++++------------
 1 file changed, 32 insertions(+), 22 deletions(-)

diff --git a/pandas/io/parsers/python_parser.py b/pandas/io/parsers/python_parser.py
index 4e1bcf54c0ae9..36d5ef7111685 100644
--- a/pandas/io/parsers/python_parser.py
+++ b/pandas/io/parsers/python_parser.py
@@ -365,6 +365,17 @@ def _convert_data(
             clean_dtypes,
         )
 
+    @cache_readonly
+    def _have_mi_columns(self) -> bool:
+        if self.header is None:
+            return False
+
+        header = self.header
+        if isinstance(header, (list, tuple, np.ndarray)):
+            return len(header) > 1
+        else:
+            return False
+
     def _infer_columns(
         self,
     ) -> tuple[list[list[Scalar | None]], int, set[Scalar | None]]:
@@ -372,18 +383,16 @@ def _infer_columns(
         num_original_columns = 0
         clear_buffer = True
         unnamed_cols: set[Scalar | None] = set()
-        self._header_line = None
 
         if self.header is not None:
             header = self.header
+            have_mi_columns = self._have_mi_columns
 
             if isinstance(header, (list, tuple, np.ndarray)):
-                have_mi_columns = len(header) > 1
                 # we have a mi columns, so read an extra line
                 if have_mi_columns:
                     header = list(header) + [header[-1] + 1]
             else:
-                have_mi_columns = False
                 header = [header]
 
             columns: list[list[Scalar | None]] = []
@@ -531,27 +540,14 @@ def _infer_columns(
                     columns, columns[0], num_original_columns
                 )
         else:
-            try:
-                line = self._buffered_line()
-
-            except StopIteration as err:
-                if not names:
-                    raise EmptyDataError("No columns to parse from file") from err
-
-                line = names[:]
-
-            # Store line, otherwise it is lost for guessing the index
-            self._header_line = line
-            ncols = len(line)
+            ncols = len(self._header_line)
             num_original_columns = ncols
 
             if not names:
                 columns = [list(range(ncols))]
-                columns = self._handle_usecols(
-                    columns, columns[0], num_original_columns
-                )
-            elif self.usecols is None or len(names) >= num_original_columns:
-                columns = self._handle_usecols([names], names, num_original_columns)
+                columns = self._handle_usecols(columns, columns[0], ncols)
+            elif self.usecols is None or len(names) >= ncols:
+                columns = self._handle_usecols([names], names, ncols)
                 num_original_columns = len(names)
             elif not callable(self.usecols) and len(names) != len(self.usecols):
                 raise ValueError(
@@ -560,12 +556,26 @@ def _infer_columns(
                 )
             else:
                 # Ignore output but set used columns.
-                self._handle_usecols([names], names, ncols)
                 columns = [names]
-                num_original_columns = ncols
+                self._handle_usecols(columns, columns[0], ncols)
 
         return columns, num_original_columns, unnamed_cols
 
+    @cache_readonly
+    def _header_line(self):
+        # Store line for reuse in _get_index_name
+        if self.header is not None:
+            return None
+
+        try:
+            line = self._buffered_line()
+        except StopIteration as err:
+            if not self.names:
+                raise EmptyDataError("No columns to parse from file") from err
+
+            line = self.names[:]
+        return line
+
     def _handle_usecols(
         self,
         columns: list[list[Scalar | None]],