fix: don't raise ArgumentError when filters map has non-integer keys

[linhyojee]

Summary

Flop.Meta.filters_to_list/1 calls String.to_integer/1 directly on every key of the filters map, raising ArgumentError on any non-integer key. This causes Flop to crash with an uncaught exception before it can return its documented {:error, %Flop.Meta{}} validation tuple.

The crash also affects Flop's own error-construction path: Flop.Meta.with_errors/3 calls filters_to_list/1 when building the params field of an error meta, so once a malformed map reaches Flop, even constructing the error tuple re-raises. That makes it impossible for callers to recover via the public Flop.validate/2 / Flop.validate_and_run/3 contract.

Repro

iex(1)> Flop.validate(%{"filters" => %{"name" => "foo"}})
** (ArgumentError) errors were found at the given arguments:

  * 1st argument: not a textual representation of an integer

    (erts 16.2) :erlang.binary_to_integer("name")
    (flop 0.26.3) lib/flop/meta.ex:181: anonymous fn/1 in Flop.Meta.filters_to_list/1
    (elixir 1.19.5) lib/enum.ex:1696: anonymous fn/3 in Enum.map/2
    (stdlib 7.2) maps.erl:894: :maps.fold_1/4
    (elixir 1.19.5) lib/enum.ex:2532: Enum.map/2
    (flop 0.26.3) lib/flop/meta.ex:181: Flop.Meta.filters_to_list/1
    (flop 0.26.3) lib/flop/meta.ex:167: Flop.Meta.with_errors/3
    (flop 0.26.3) lib/flop.ex:1382: Flop.validate/2
    iex:1: (file)
iex(1)> 

Observed in production from a misconfigured client sending ?filters[transport_type_code]=SEA. The map-keyed-by-field name form is invalid input, but it should surface as a validation error rather than an uncaught exception.

Change

Replace String.to_integer/1 with a parse_filter_index/1 helper that accepts integer keys as-is and uses Integer.parse/1 for string keys, requiring the entire string to parse. Entries with non-integer keys are dropped silently from the resulting list - matching how the fallthrough clause already treats filters values that aren't in the indexed-map shape. Behaviour for all currently-valid inputs is unchanged.

Tests

Added in test/base/flop/meta_test.exs:

• Regression: filters are sorted by numeric index, not lexicographic
• New: non-integer keys are dropped, mixed maps keep only valid entries
• New: field-name-keyed map no longer raises
• Regression: list-form filters pass through untouched
• Round-trip: Flop.validate/2 returns {:error, %Flop.Meta{}} for a field-name-keyed filter map

[woylie]

Wouldn't you want to see the original parameters in the error struct?

[linhyojee]

Good point 👍

filters_to_list/1 is only called from with_errors/3, so its output ends up in meta.params. Silently dropping malformed entries hides exactly the information the caller needs to debug (e.g, "here's the filters[name]=foo you sent, here's why it's wrong"...etc).

Updated (commit: 1ad0d9f): when any key can't be parsed as integer, leave the input filters untouched rather than partially transforming it. The happy path (all integer indexed keys) unchanged, and a malformed map goes into meta.params so the caller sees exactly what they sent.

[linhyojee]

Hi @woylie Are you happy with the update/change?

[woylie]

Looks good, thank you!

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.41%. Comparing base (e7f04ae) to head (1ad0d9f).
⚠️ Report is 14 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/flop/meta.ex	85.71%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #655      +/-   ##
==========================================
- Coverage   84.45%   84.41%   -0.05%     
==========================================
  Files          15       15              
  Lines        1023     1033      +10     
==========================================
+ Hits          864      872       +8     
- Misses        159      161       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.