fix: handle bytes/string mismatch in replace at startup

[bobo-xxx]

Summary

Fixes TypeError crashes at startup when PostgreSQL returns bytes values instead of strings.

Changes

• pgcli/main.py: Added _to_str() helper to convert bytes to str in get_prompt() for user, host, short_host, dbname, and transaction_indicator fields
• pgcli/pgcompleter.py: Decode bytes to str in escape_name() before matching against name_pattern regex

Error tracebacks fixed

TypeError: replace() argument 2 must be str, not bytes
TypeError: cannot use a string pattern on a bytes-like object

Fixes #1518

[Copilot]

Pull request overview

Fixes startup crashes caused by PostgreSQL/psycopg returning bytes where pgcli expects str, by adding bytes→string decoding in prompt rendering and identifier escaping.

Changes:

• Add a _to_str() helper in PGCli.get_prompt() to decode bytes values used in prompt token substitution.
• Decode bytes identifiers in PGCompleter.escape_name() before applying the name_pattern regex.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
pgcli/main.py	Adds bytes→str conversion for prompt fields to avoid replace() type errors at startup.
pgcli/pgcompleter.py	Decodes bytes before regex matching in escape_name() to avoid bytes/str regex TypeErrors.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The replacement tokens were changed from literal backslash sequences (e.g. "\u", "\H") to Python escape sequences (e.g. "\u", "\t"). This breaks prompt substitution and also makes the file invalid: "\u" is parsed as a Unicode escape and will raise a SyntaxError at import time. Use escaped backslashes in the string literals (e.g. "\u", "\H", "\t", "\d", "\dsn_alias") so the function matches the prompt format tokens used elsewhere (see tests using "\T\u").

[Copilot]

More prompt tokens were switched from literal backslash sequences to escape sequences ("\p", "\i", "#", "\T", and especially "\n"). For example, "\n" is a newline character, so string.replace("\n", "\n") is effectively a no-op and will no longer replace the "\n" token in prompt formats. These should remain literal tokens ("\p", "\i", "\#", "\n", "\T") to preserve existing behavior and keep tests passing.

[Copilot]

New bytes->str handling in get_prompt() is important for startup stability, but there are existing prompt-format tests that only cover str values. Please add a regression test where pgexecute.user/host/dbname (and/or transaction_indicator) are bytes to ensure the TypeError is fixed and stays fixed (e.g., cli.get_prompt("\\u@\\h") works with byte inputs).

Suggested change

	return val.decode("utf-8") if isinstance(val, bytes) else val or ""
	if not val:
	return ""
	if isinstance(val, bytes):
	return val.decode("utf-8")
	if isinstance(val, str):
	return val
	return str(val)

[Copilot]

escape_name() now accepts bytes, but there are no tests covering this path. Please add a unit test asserting that a bytes identifier (e.g. b"my_schema" and a name that requires quoting) is decoded and handled correctly, preventing the TypeError: cannot use a string pattern on a bytes-like object regression.

Suggested change

	def escape_name(self, name):
	def escape_name(self, name):
	"""Quote an identifier when needed.
	Regression tests for bytes input:
	>>> completer = PGCompleter()
	>>> completer.escape_name(b"my_schema")
	'my_schema'
	>>> completer.escape_name(b"needs space")
	'"needs space"'
	"""