Summary

After capitalize() is called on a name that has no middle name (or any other empty attribute), surnames carries a spurious leading space and the *_list attributes become [''] instead of [].

Reproduction

from nameparser import HumanName

n = HumanName('john doe')
n.capitalize()

print(repr(n.surnames))     # ' Doe'   <- leading space
print(n.middle_list)        # ['']    <- should be []
print(n.title_list)         # ['']    <- should be []

The leading space is a real observable artifact, not just an empty-list detail: downstream code that joins or compares surnames silently breaks.

Cause

cap_piece(...).split(' ') uses str.split(' ') with an explicit separator. On an empty string, split(' ') returns [''] (it preserves the empty token), whereas str.split() with no argument collapses runs of whitespace and returns [] for an empty/whitespace-only string — which is the correct result for a list of name tokens.

Fix

Drop the explicit separator for the whitespace-delimited attributes:

-        self.title_list = self.cap_piece(self.title, 'title').split(' ')
-        self.first_list = self.cap_piece(self.first, 'first').split(' ')
-        self.middle_list = self.cap_piece(self.middle, 'middle').split(' ')
-        self.last_list = self.cap_piece(self.last, 'last').split(' ')
+        self.title_list = self.cap_piece(self.title, 'title').split()
+        self.first_list = self.cap_piece(self.first, 'first').split()
+        self.middle_list = self.cap_piece(self.middle, 'middle').split()
+        self.last_list = self.cap_piece(self.last, 'last').split()
         self.suffix_list = self.cap_piece(self.suffix, 'suffix').split(', ')

The suffix line is intentionally left as .split(', '), because commas are a deliberate delimiter there.

Tests

Adds test_capitalize_empty_name_part_has_no_leading_space_in_surnames, asserting surnames == 'Doe', middle_list == [], and surnames_list == ['Doe'] after capitalize() (covers the force=True path too). The case fails on the current code and passes with the fix. Full suite: 345 tests OK (skipped=2, expected failures=10).

I couldn't find an existing issue or PR covering this — happy to fold it into one if I missed it.