fix: strip underscore prefix from digit-starting codes in all serialization paths

[ktrufanov-skycop]

Summary

Follow-up to #90 (merged). Addresses the serialization gap flagged
by @punitarani in the review comment —
digit-prefixed airline codes like _3F were still leaking into API
requests and CLI output through paths not covered by the first PR.

Changes

• removeprefix("_") in flights.py and dates.py serialize()
  functions — strip the synthetic underscore before sending codes to
  Google Flights API
• serialize_airline() in cli/utils.py — same fix for CLI JSON output
• Round-trip selected_flight in flights.py — pass enum objects
  (leg.airline, leg.departure_airport, leg.arrival_airport) to
  serialize() instead of pre-resolving .name, which bypassed the
  isinstance check and leaked _3F into the request payload
• Test for serialize_airline() with numeric-prefix code

Why removeprefix over lstrip

lstrip("_") would strip all leading underscores (__3F → 3F).
removeprefix("_") removes exactly one — matching the single underscore
added by generate_enums.py for digit-starting IATA codes.

Test plan

• serialize_airline(Airline._3F) returns {"code": "3F", "name": "FlyOne Armenia"}
• All 6 airline-related tests pass
• No regressions in existing tests

Greptile Summary

This PR completes the fix for digit-prefixed IATA airline codes (e.g., _3F) leaking out of serialization paths, following up on #90. It applies removeprefix(\"_\") in the two Google Flights API serializers (FlightSearchFilters and DateSearchFilters), in the CLI JSON output helper (serialize_airline), and — critically — fixes the selected_flight round-trip block in flights.py that was passing .name strings directly to serialize(), bypassing the isinstance(obj, Airline) branch entirely.

• fli/models/google_flights/flights.py: removeprefix(\"_\") added to serialize(), plus leg.departure_airport, leg.arrival_airport, and leg.airline now passed as enum objects (not pre-resolved .name strings) in the selected_flights builder — this was the root cause of the payload leak.
• fli/models/google_flights/dates.py: Same removeprefix(\"_\") fix applied to the parallel serialize() in DateSearchFilters.format().
• fli/cli/utils.py: serialize_airline() now strips the underscore from the code key in CLI JSON output.
• tests/cli/test_utils.py: New test_serialize_airline_numeric_prefix test verifies Airline._3F → {\"code\": \"3F\", \"name\": \"FlyOne Armenia\"}.
• Minor inconsistency: fli/mcp/server.py line 294 still uses lstrip(\"_\") for airline_code, which differs from the removeprefix(\"_\") approach used everywhere else after this PR. This is a pre-existing gap not introduced here, but worth aligning.

Confidence Score: 5/5

Safe to merge — all serialization paths that send data to the Google Flights API or emit CLI/JSON output are correctly fixed.

All changed code is correct. The critical root-cause fix (passing enum objects instead of pre-resolved .name strings in the selected_flights block) is sound. The only remaining gap is a pre-existing lstrip("_") in the MCP server that is not introduced by this PR and is P2 only.

fli/mcp/server.py line 294 — pre-existing lstrip("_") inconsistency worth aligning with the removeprefix("_") approach used everywhere else after this PR.

Important Files Changed

Filename	Overview
fli/models/google_flights/flights.py	Two fixes: removeprefix("_") in serialize() for Airport/Airline enums, and passing enum objects (not .name strings) to serialize() in the selected_flights block — closing the bypass of the isinstance check that leaked _3F into API payloads.
fli/models/google_flights/dates.py	Same removeprefix("_") fix applied to the parallel serialize() function inside DateSearchFilters.format(); no selected-flight block in this model so no further changes needed.
fli/cli/utils.py	serialize_airline() updated to strip the synthetic underscore from the code key in CLI JSON output; serialize_airport() is unchanged (correct, as Airport has no digit-starting codes).
tests/cli/test_utils.py	New test_serialize_airline_numeric_prefix test verifies that Airline._3F round-trips to {"code": "3F", "name": "FlyOne Armenia"} via serialize_airline().

Sequence Diagram

sequenceDiagram
    participant User
    participant Serializer
    participant GoogleFlightsAPI

    Note over User,GoogleFlightsAPI: Before PR — digit-prefix leak
    User->>Serializer: leg.airline.name passes string "_3F"
    Note over Serializer: No isinstance match, falls through
    Serializer->>GoogleFlightsAPI: payload contains "_3F" (incorrect)

    Note over User,GoogleFlightsAPI: After PR — correct round-trip
    User->>Serializer: leg.airline passes Airline enum object
    Note over Serializer: isinstance(obj, Airline) matches
    Serializer->>GoogleFlightsAPI: obj.name.removeprefix returns "3F" (correct)

Loading

Comments Outside Diff (1)

1. fli/mcp/server.py, line 294 (link)

   Inconsistent strip method vs. the rest of the codebase

   The MCP server still uses lstrip("_") here, while every other serialization path in this PR (and the PR description itself) explicitly prefers removeprefix("_") — because lstrip("_") would strip all leading underscores (e.g., a hypothetical __3F → 3F), whereas removeprefix("_") removes exactly the one underscore added by generate_enums.py.

   For parity with the rest of the codebase after this PR:

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: fli/mcp/server.py
   Line: 294
   
   Comment:
   **Inconsistent strip method vs. the rest of the codebase**
   
   The MCP server still uses `lstrip("_")` here, while every other serialization path in this PR (and the PR description itself) explicitly prefers `removeprefix("_")` — because `lstrip("_")` would strip *all* leading underscores (e.g., a hypothetical `__3F` → `3F`), whereas `removeprefix("_")` removes exactly the one underscore added by `generate_enums.py`.
   
   For parity with the rest of the codebase after this PR:
   
   
   
   How can I resolve this? If you propose a fix, please make it concise.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: fli/mcp/server.py
Line: 294

Comment:
**Inconsistent strip method vs. the rest of the codebase**

The MCP server still uses `lstrip("_")` here, while every other serialization path in this PR (and the PR description itself) explicitly prefers `removeprefix("_")` — because `lstrip("_")` would strip *all* leading underscores (e.g., a hypothetical `__3F` → `3F`), whereas `removeprefix("_")` removes exactly the one underscore added by `generate_enums.py`.

For parity with the rest of the codebase after this PR:

```suggestion
        "airline_code": getattr(leg.airline, "name", str(leg.airline)).removeprefix("_"),
```

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "fix: strip underscore prefix from digit-..." | Re-trigger Greptile}

_{(4/5) You can add custom instructions or style guidelines for the agent here!}

[ktrufanov-skycop]

Duplicate — resolved conflict in #94 instead.