Implementing Tab Completion

Python
import readline
import rlcompleter

readline.parse_and_bind("tab: complete")

That's it for the standard interactive interpreter case—rlcompleter.Completer binds against the __main__ namespace by default when used this way (as python -m startup does via site.py / PYTHONSTARTUP).

Progress:

• Step 1: Decide the namespace to complete against (global __main__ vs. a custom dict/namespace)
• Step 2: Instantiate rlcompleter.Completer (optionally with a namespace)
• Step 3: Register the completer with readline.set_completer
• Step 4: Bind the Tab key via readline.parse_and_bind
• Step 5: Test simple identifier completion and dotted attribute completion

Step 1–4: Custom namespace example

Python
import readline
import rlcompleter

namespace = {"foo": 42, "bar": "hello"}
completer = rlcompleter.Completer(namespace)

readline.set_completer(completer.complete)
readline.parse_and_bind("tab: complete")

Step 5: How completion resolves

• complete(text, state) is called repeatedly by readline with increasing state (0, 1, 2...) until it returns None.
• No dot in text → matches keywords, builtins, and names in the namespace.
• Dot in text (e.g. "foo.up") → evaluates the expression before the last dot using eval() in the namespace, then completes attributes via dir(). This means completion can execute arbitrary code (property getters, __getattr__, etc.).

Example 1: Default interactive interpreter setup Input: Add tab completion to a script meant to be run with python -i myscript.py Output:

Python
try:
    import readline
except ImportError:
    pass
else:
    import rlcompleter
    readline.parse_and_bind("tab: complete")

Example 2: Completion inside a sandboxed REPL Input: Build a mini-REPL that only completes names from a restricted namespace, not __main__ Output:

Python
import readline, rlcompleter

sandbox_ns = {"__builtins__": {}, "x": [1, 2, 3]}
completer = rlcompleter.Completer(sandbox_ns)
readline.set_completer(completer.complete)
readline.parse_and_bind("tab: complete")

while True:
    line = input(">>> ")
    exec(line, sandbox_ns)

Example 3: Inspecting raw completion matches (non-interactive/testing) Input: Programmatically get completions for "li" and "str.up" Output:

Python
c = rlcompleter.Completer({"list": list})
i = 0
matches = []
while True:
    m = c.complete("li", i)
    if m is None:
        break
    matches.append(m)
    i += 1
# matches -> ['list(', 'list.']  (exact strings vary by Python version)

i, attr_matches = 0, []
while True:
    m = c.complete("str.up", i)
    if m is None:
        break
    attr_matches.append(m)
    i += 1
# attr_matches -> ['str.upper(']