`_search` drops results from successful backends when another backend raises first

[ddett]

Before you open an issue:

• I have the latest version. (Check: ddgs version. Update: pip install -U ddgs)
• I tried reinstalling the library. (pip install -I ddgs)
• I tried using a proxy

Describe the bug

When text() is called with multiple backends (e.g. backend="brave,google,duckduckgo"), a single failing backend can cause all results — including those from successful backends — to be silently dropped, raising DDGSException: No results found.

The issue is in _search() (ddgs.py L191–216):

done, not_done = wait(futures.values(), return_when=FIRST_EXCEPTION)

If a backend raises before the others complete, the successful futures land in not_done and are never collected. Only done is iterated — which contains the failed future — so err is set, and the method raises instead of returning the results that other backends would have produced.

Steps to reproduce the problem:

from ddgs import DDGS

# google backend reliably fails (see #412); if it raises before
# brave/duckduckgo finish, their results are silently discarded
results = DDGS().text("python programming", backend="brave,google,duckduckgo", max_results=10)
# -> DDGSException: No results found.

# Same query without the failing backend works fine:
results = DDGS().text("python programming", backend="brave,duckduckgo", max_results=10)
# -> returns 10 results

The bug is non-deterministic: because backends race, sometimes a healthy backend finishes before the failing one raises, and results appear. Repeat the first snippet a few times and it will fail intermittently.

Expected behavior: If any backend returns results, those should be included. Only raise if every backend fails.

Suggested fix: Replace FIRST_EXCEPTION with ALL_COMPLETED (or as_completed) and collect results from all futures that succeeded:

done, _ = wait(futures.values(), return_when=ALL_COMPLETED)
results = []
for future in done:
    try:
        results.extend(future.result())
    except Exception:
        pass  # skip failed backends
if not results:
    raise DDGSException("No results found.")

Specify this information

• OS: macOS 15 / Ubuntu 22.04
• Environment: Python 3.12, pip venv
• ddgs version: 9.11.3

Related issues:

• The issue of backend choosing duckduckgo #390 — same root cause (single duckduckgo backend fails due to rate limiting; adding more backends "fixes" it only because they sometimes return before the exception)
• No results with backend = google #412 — google backend failures would trigger this when combined with other backends

[deedy5]

Take a closer look at what's happening here:

ddgs/ddgs/ddgs.py

Lines 174 to 202 in aa229fd

	for i, engine in enumerate(engines, start=1):
	if engine.provider in seen_providers:
	continue
	future = executor.submit(
	engine.search,
	query,
	region=region,
	safesearch=safesearch,
	timelimit=timelimit,
	page=page,
	**kwargs,
	)
	futures[future] = engine
	if len(futures) >= max_workers or i >= max_workers:
	done, not_done = wait(futures, timeout=self._timeout, return_when="FIRST_EXCEPTION")
	for f, f_engine in futures.items():
	if f in done:
	try:
	if r := f.result():
	results_aggregator.extend(r)
	seen_providers.add(f_engine.provider)
	except Exception as ex: # noqa: BLE001
	err = ex
	logger.info("Error in engine %s: %r", engine.name, ex)
	futures = {f: futures[f] for f in not_done}
	if max_results and len(results_aggregator) >= max_results:
	break

[unixfox]

I can reproduce the same issue. When doing many search in a row, you can replicate the issue fairly easily.

[deedy5]

I can't reproduce the problem.