qubit-json 0.3.1

/*******************************************************************************
 *
 *    Copyright (c) 2026 Haixing Hu.
 *
 *    SPDX-License-Identifier: Apache-2.0
 *
 *    Licensed under the Apache License, Version 2.0.
 *
 ******************************************************************************/
//! Tests for normalization behavior implemented in `lenient_json_normalizer.rs`.
//!

use serde_json::json;

use qubit_json::{JsonDecodeErrorKind, JsonDecodeOptions, JsonDecodeStage, LenientJsonDecoder};

#[test]
fn test_decode_value_reports_empty_input_for_empty_string() {
    let decoder = LenientJsonDecoder::default();
    let error = decoder
        .decode_value("")
        .expect_err("empty input should be rejected before JSON parsing");
    assert_eq!(error.kind, JsonDecodeErrorKind::EmptyInput);
}

#[test]
fn test_decode_value_reports_empty_input_for_whitespace_by_default() {
    let decoder = LenientJsonDecoder::default();
    let error = decoder
        .decode_value(" \n\t ")
        .expect_err("whitespace-only input should be empty after default trimming");
    assert_eq!(error.kind, JsonDecodeErrorKind::EmptyInput);
}

#[test]
fn test_decode_value_reports_invalid_json_for_whitespace_when_trimming_disabled() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        trim_whitespace: false,
        strip_markdown_code_fence: false,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("   ")
        .expect_err("whitespace-only input should reach JSON parser when trimming is disabled");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_respects_input_size_limit() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        max_input_bytes: Some(6),
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("{\"a\":1}")
        .expect_err("input above the configured byte limit should be rejected");
    assert_eq!(error.kind, JsonDecodeErrorKind::InputTooLarge);
    assert_eq!(error.stage, JsonDecodeStage::Normalize);
    assert_eq!(error.input_bytes, Some(7));
    assert_eq!(error.max_input_bytes, Some(6));
    assert!(error.to_string().contains("6 bytes"));
}

#[test]
fn test_decode_value_accepts_input_at_size_limit() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        max_input_bytes: Some(7),
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("[1,2,3]")
        .expect("input whose size matches the limit should be accepted");
    assert_eq!(value, json!([1, 2, 3]));
}

#[test]
fn test_decode_value_size_limit_runs_before_parser_error_mapping() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        max_input_bytes: Some(0),
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("{")
        .expect_err("size guard should run before parser handling");
    assert_eq!(error.kind, JsonDecodeErrorKind::InputTooLarge);
}

#[test]
fn test_decode_value_strips_utf8_bom_by_default() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("\u{feff}{\"a\":1}")
        .expect("default decoder should strip a leading UTF-8 BOM");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_reports_empty_input_when_only_bom_is_present() {
    let decoder = LenientJsonDecoder::default();
    let error = decoder
        .decode_value("\u{feff}")
        .expect_err("input containing only BOM should become empty after normalization");
    assert_eq!(error.kind, JsonDecodeErrorKind::EmptyInput);
}

#[test]
fn test_decode_value_can_leave_utf8_bom_when_disabled() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_utf8_bom: false,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("\u{feff}{\"a\":1}")
        .expect_err("BOM should remain and break parsing when BOM stripping is disabled");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_strips_code_fence_with_closing_fence() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("```json\n{\"a\":1}\n```")
        .expect("default decoder should strip a closing Markdown code fence");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_strips_tilde_code_fence() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("~~~json\n{\"a\":1}\n~~~")
        .expect("default decoder should strip a tilde Markdown code fence");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_strips_indented_code_fence_when_trimming_disabled() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        trim_whitespace: false,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("  ```json\n{\"a\":1}\n  ```")
        .expect("decoder should accept up to three leading spaces before a fence");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_rejects_deeply_indented_code_fence_when_trimming_disabled() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        trim_whitespace: false,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("    ```json\n{\"a\":1}\n    ```")
        .expect_err("deeply indented fences should remain ordinary text");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_strips_code_fence_with_more_than_three_backticks() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("````json\n{\"text\":\"```\"}\n````")
        .expect("decoder should strip matching Markdown fences longer than three backticks");
    assert_eq!(value, json!({"text": "```"}));
}

#[test]
fn test_decode_value_strips_code_fence_with_longer_closing_fence() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("```json\n{\"a\":1}\n````")
        .expect("decoder should accept a closing fence longer than the opening fence");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_strips_code_fence_with_indented_closing_fence() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("```json\n{\"a\":1}\n   ```   \n")
        .expect("decoder should accept a closing fence alone on a whitespace-padded line");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_rejects_closing_fence_shorter_than_opening_fence() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_requires_closing: true,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("````json\n{\"a\":1}\n```")
        .expect_err("closing fence shorter than the opening fence should not be stripped");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_strips_code_fence_without_closing_fence() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("```json\n{\"a\":1}")
        .expect("default decoder should strip an opening fence even without a closing fence");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_can_require_closing_code_fence() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_requires_closing: true,
        ..JsonDecodeOptions::default()
    });
    let error = decoder.decode_value("```json\n{\"a\":1}").expect_err(
        "opening fence without closing fence should be rejected when strict mode is enabled",
    );
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_allows_strict_closing_code_fence_when_present() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_requires_closing: true,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("```json\n{\"a\":1}\n```")
        .expect("strict closing mode should still strip a properly closed fence");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_can_restrict_code_fence_to_json_language_tags() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_json_only: true,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("```python\n{\"a\":1}\n```")
        .expect_err("non-JSON code fence should not be stripped in json-only mode");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_json_only_mode_accepts_longer_json_code_fence() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_json_only: true,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("````JSON\n{\"a\":1}\n````")
        .expect("json-only mode should accept longer JSON fenced blocks");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_json_only_mode_accepts_jsonc_code_fence() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_json_only: true,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("```jsonc\n{\"a\":1}\n```")
        .expect("json-only mode should accept jsonc fenced blocks");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_json_only_mode_accepts_empty_code_fence_tag() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_json_only: true,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("```\n{\"a\":1}\n```")
        .expect("json-only mode should accept fenced blocks without a language tag");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_json_only_mode_accepts_json_info_string() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_json_only: true,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("```json title=\"sample\"\n{\"a\":1}\n```")
        .expect("json-only mode should accept JSON fenced blocks with info strings");
    assert_eq!(value, json!({"a": 1}));
}

#[test]
fn test_decode_value_json_only_mode_rejects_non_json_info_string_first_token() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence_json_only: true,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("```python json\n{\"a\":1}\n```")
        .expect_err("json-only mode should use the first info-string token");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_does_not_accept_inline_closing_ticks_as_fence_end() {
    let decoder = LenientJsonDecoder::default();
    let error = decoder
        .decode_value("```json\n{\"a\":1}```")
        .expect_err("inline trailing ticks are not treated as a valid closing fence");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_randomized_inputs_do_not_panic_and_round_trip_when_valid() {
    let decoders = [
        LenientJsonDecoder::default(),
        LenientJsonDecoder::new(JsonDecodeOptions {
            trim_whitespace: false,
            strip_markdown_code_fence: false,
            ..JsonDecodeOptions::default()
        }),
        LenientJsonDecoder::new(JsonDecodeOptions {
            strip_markdown_code_fence_json_only: true,
            strip_markdown_code_fence_requires_closing: true,
            ..JsonDecodeOptions::default()
        }),
    ];

    let mut seed = 0x0d15_ea5e_d5e0_ded5u64;
    for _ in 0..3000 {
        let input = generate_noisy_json_candidate(&mut seed);
        for decoder in &decoders {
            let result = std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| {
                decoder.decode_value(&input)
            }));
            assert!(result.is_ok(), "decoder panicked on input: {input:?}");

            if let Ok(value) = result.expect("catch_unwind returned no result") {
                let canonical = serde_json::to_string(&value)
                    .expect("serializing a decoded JSON value should not fail");
                let reparsed = decoder
                    .decode_value(&canonical)
                    .expect("canonical JSON should be decodable by the same decoder");
                assert_eq!(reparsed, value);
            }
        }
    }
}

fn generate_noisy_json_candidate(seed: &mut u64) -> String {
    const ALPHABET: &[char] = &[
        '{', '}', '[', ']', ':', ',', '"', '\\', '`', ' ', '\t', '\n', '\r', 'a', 'b', 'c', 'x',
        'y', 'z', '0', '1', '2', '9', '-', '.', 't', 'f', 'n', '\u{0000}', '\u{0008}', '\u{001f}',
    ];

    let len = (next_u64(seed) % 48) as usize;
    let mut text = String::with_capacity(len + 16);
    for _ in 0..len {
        let index = (next_u64(seed) % ALPHABET.len() as u64) as usize;
        text.push(ALPHABET[index]);
    }

    match next_u64(seed) % 4 {
        0 => format!("```json\n{text}\n```"),
        1 => format!("```python\n{text}\n```"),
        2 => format!("\u{feff}{text}"),
        _ => text,
    }
}

fn next_u64(seed: &mut u64) -> u64 {
    *seed = seed
        .wrapping_mul(6364136223846793005)
        .wrapping_add(1442695040888963407);
    *seed
}

#[test]
fn test_decode_value_reports_invalid_json_for_code_fence_without_newline() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        trim_whitespace: false,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("```json")
        .expect_err("text without a fence body newline should not be stripped");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_reports_empty_input_for_empty_code_fence_body() {
    let decoder = LenientJsonDecoder::default();
    let error = decoder
        .decode_value("```json\n```")
        .expect_err("empty fenced body should become empty input after normalization");
    assert_eq!(error.kind, JsonDecodeErrorKind::EmptyInput);
}

#[test]
fn test_decode_value_can_disable_code_fence_stripping() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        strip_markdown_code_fence: false,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("```json\n{\"name\":\"alice\"}\n```")
        .expect_err("code fences should remain when stripping is disabled");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_preserves_existing_escapes() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("{\"text\":\"a\\nb\"}")
        .expect("existing JSON escapes should remain valid");
    assert_eq!(value, json!({"text": "a\nb"}));
}

#[test]
fn test_decode_value_escapes_control_chars_in_strings() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("{\"text\":\"a\nb\"}")
        .expect("default decoder should escape control characters inside strings");
    assert_eq!(value, json!({"text": "a\nb"}));
}

#[test]
fn test_decode_value_can_disable_control_char_escaping() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        escape_control_chars_in_strings: false,
        ..JsonDecodeOptions::default()
    });
    let error = decoder
        .decode_value("{\"text\":\"a\nb\"}")
        .expect_err("control characters should remain invalid JSON when escaping is disabled");
    assert_eq!(error.kind, JsonDecodeErrorKind::InvalidJson);
}

#[test]
fn test_decode_value_covers_all_supported_control_char_escapes() {
    let control_chars = [
        '\u{0000}', '\u{0001}', '\u{0002}', '\u{0003}', '\u{0004}', '\u{0005}', '\u{0006}',
        '\u{0007}', '\u{0008}', '\u{0009}', '\u{000a}', '\u{000b}', '\u{000c}', '\u{000d}',
        '\u{000e}', '\u{000f}', '\u{0010}', '\u{0011}', '\u{0012}', '\u{0013}', '\u{0014}',
        '\u{0015}', '\u{0016}', '\u{0017}', '\u{0018}', '\u{0019}', '\u{001a}', '\u{001b}',
        '\u{001c}', '\u{001d}', '\u{001e}', '\u{001f}',
    ];
    let control_text: String = control_chars.into_iter().collect();
    let json_input = format!("{{\"text\":\"{control_text}\"}}");

    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value(&json_input)
        .expect("all supported ASCII control characters should be escaped successfully");
    assert_eq!(value, json!({"text": control_text}));
}

#[test]
fn test_decode_value_trims_surrounding_whitespace_by_default() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder.decode_value("\n{\"text\":\"abc\"}\n").expect(
        "leading and trailing control characters outside strings should be trimmed by default",
    );
    assert_eq!(value, json!({"text": "abc"}));
}

#[test]
fn test_decode_value_with_trim_disabled_and_escape_enabled_still_decodes_owned_output() {
    let decoder = LenientJsonDecoder::new(JsonDecodeOptions {
        trim_whitespace: false,
        strip_markdown_code_fence: false,
        escape_control_chars_in_strings: true,
        ..JsonDecodeOptions::default()
    });
    let value = decoder
        .decode_value("{\"text\":\"a\nb\"}")
        .expect("escaping inside strings should still work when trimming is disabled");
    assert_eq!(value, json!({"text": "a\nb"}));
}

#[test]
fn test_decode_value_trims_owned_output_after_repair() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("```json\n  {\"text\":\"a\nb\"}  \n```")
        .expect("trim should still apply after repair allocates an owned string");
    assert_eq!(value, json!({"text": "a\nb"}));
}

#[test]
fn test_decode_value_handles_uppercase_code_fence_language_tag() {
    let decoder = LenientJsonDecoder::default();
    let value = decoder
        .decode_value("```JSON\n{\"a\":1}\n```")
        .expect("code fence stripping should not depend on the language tag case");
    assert_eq!(value, json!({"a": 1}));
}