README.md

Chapter 4 — Unicode Text Versus Bytes

Chapter goal: Internalize the bytes/str divide, understand how Python handles encoding errors, and be able to write platform-safe text I/O — without thinking about it.

---

What You'll Learn

Section	Core concept
Character Issues	Code points vs byte representations
Byte Essentials	bytes, bytearray, memoryview
Basic Codecs	UTF-8, UTF-16, latin-1, cp1252
Encode/Decode Errors	Four error handlers, UnicodeEncodeError vs UnicodeDecodeError
Text File Handling	Unicode sandwich, always-explicit encoding=
Encoding Defaults	Why Windows breaks "works on my machine" code
Normalization	NFC, NFD, NFKC, NFKD — canonical equivalents
Case Folding	casefold() vs lower() for non-ASCII

---

Key Files

• examples.py — 8-part annotated implementation of all major concepts
• exercises.py — 7 original exercises pushing at edge cases
• benchmarks.py — UTF-8 vs UTF-32 memory; encode/decode overhead; normalization cost
• mini_project.py — Unicode-safe CSV cleaner and text normalizer
• notes.md — Deep technical reference with tables and diagrams
• pitfalls.md — 7 common encoding bugs with root causes and fixes
• interview_questions.md — 9 questions from L3→L6
• architecture_notes.md — Why Unicode was designed the way it was

---

The Central Mental Model

         ENCODE                         DECODE
str  ─────────────────→  bytes  ────────────────────→  str
(code points)          (raw bytes)               (code points)
(abstract)             (concrete)                (abstract)
(in memory)            (on disk/network)         (in memory)

"café"  → utf-8 →  b'caf\xc3\xa9'  → utf-8 →  "café"
  4 chars             5 bytes                    4 chars

encode = "serialize for machines"
decode = "interpret for humans"

---

The str/bytes Separation — Why It Matters

Python 2 blurred the line between text and bytes. This caused:

• Silent mojibake (garbled text) instead of explicit errors
• Code that worked on ASCII input but silently corrupted international data
• No way to know whether a string contained text or raw bytes

Python 3 makes the divide explicit. The cost: you must think about encoding at boundaries. The benefit: encoding bugs become immediately visible as TypeErrors instead of appearing as corrupted data downstream.

---

Encoding Error Handler Reference

city = "São Paulo"

city.encode("cp437")                           # UnicodeEncodeError (default: 'strict')
city.encode("cp437", errors="ignore")          # b'So Paulo'  ← data loss!
city.encode("cp437", errors="replace")         # b'S?o Paulo' ← lossy
city.encode("cp437", errors="xmlcharrefreplace")  # b'São Paulo' ← safe for XML

---

Normalization Quick Reference

from unicodedata import normalize

s1 = "café"                              # U+00E9 (composed é)
s2 = "cafe\N{COMBINING ACUTE ACCENT}"   # e + U+0301 (decomposed)

# Both forms, visually identical, compare as unequal:
s1 == s2  # False — different code point sequences

# Normalize before comparing:
normalize("NFC", s1) == normalize("NFC", s2)  # True
normalize("NFD", s1) == normalize("NFD", s2)  # True

# Diacritic removal (for fuzzy search):
nfd = normalize("NFD", s1)
"".join(c for c in nfd if unicodedata.category(c) != "Mn")  # "cafe"

---

The Platform Encoding Trap — Reference

Setting	Linux/macOS	Windows (typical)
locale.getpreferredencoding()	'UTF-8'	'cp1252'
File open default	'UTF-8'	'cp1252'
sys.stdout.encoding (interactive)	'utf-8'	'utf-8' (since 3.6)
sys.stdout.encoding (redirected)	'UTF-8'	'cp1252'
sys.getfilesystemencoding()	'utf-8'	'utf-8' (since 3.6)

Fix everything at once: sys.stdout.reconfigure(encoding="utf-8") at the top of scripts.

---

Pre-Reading Checklist

Before working through this chapter, make sure you understand:

• Python's type system distinguishes str and bytes absolutely
• "café"[1] → "a" (a str); b"caf\xc3\xa9"[3] → 195 (an int!)
• len("café") → 4; len("café".encode("utf-8")) → 5

If any of those surprised you, start with examples.py Part 1.

---

Reference: Fluent Python 2nd ed., Chapter 4 — pages 117–165