trustformers 0.1.1

#![allow(clippy::all)]
use proptest::prelude::*;
use std::collections::{HashMap, HashSet};
use trustformers::core::tensor::Tensor;
use trustformers::tokenizers::TokenizerWrapper;
use trustformers::*;

/// Fuzzing tests for TrustformeRS components
/// Tests robustness against unexpected inputs and edge cases

#[cfg(test)]
mod fuzz_tests {
    use super::*;

    /// Fuzz test for tensor operations
    #[test]
    fn fuzz_tensor_operations() {
        fn tensor_ops_test(
            shape: Vec<usize>,
            data: Vec<f32>,
        ) -> std::result::Result<(), TestCaseError> {
            if shape.is_empty() || shape.contains(&0) {
                return Ok(()); // Skip invalid shapes
            }

            let total_elements: usize = shape.iter().product();
            if total_elements > 1_000_000 || data.len() != total_elements {
                return Ok(()); // Skip extremely large tensors or mismatched data
            }

            let tensor = Tensor::from_slice(&data, &shape);

            match tensor {
                Ok(t) => {
                    // Test basic operations don't panic
                    let _ = t.shape();
                    let _ = t.data();

                    // Test mathematical operations
                    if let Ok(zeros) = Tensor::zeros(&shape) {
                        let _ = t.add(&zeros);
                        let _ = t.mul(&zeros);
                    }

                    // Test reshaping
                    if total_elements > 1 {
                        let _ = t.reshape(&[total_elements]);
                    }
                },
                Err(_) => {
                    // Expected for invalid inputs
                },
            }

            Ok(())
        }

        proptest!(|(
            shape in prop::collection::vec(1usize..100, 1..4),
            data_multiplier in 0.0f32..1000.0f32,
        )| {
            let total_elements: usize = shape.iter().product();
            let data: Vec<f32> = (0..total_elements)
                .map(|i| (i as f32) * data_multiplier / 1000.0)
                .collect();

            tensor_ops_test(shape, data)?;
        });
    }

    /// Fuzz test for tokenizer with random text inputs
    #[test]
    fn fuzz_tokenizer_random_text() {
        fn tokenize_test(text: String) -> std::result::Result<(), TestCaseError> {
            // Create a simple tokenizer for testing
            let vocab = HashMap::new();
            let merges = Vec::new();
            let bpe_tokenizer = trustformers::tokenizers::bpe::BPETokenizer::new(vocab, merges);
            let tokenizer = TokenizerWrapper::BPE(bpe_tokenizer);

            // Test tokenization doesn't panic on arbitrary text
            match tokenizer.encode(&text) {
                Ok(tokens) => {
                    // Verify tokens are valid
                    prop_assert!(tokens.input_ids.len() <= text.len() * 2 + 10); // Reasonable upper bound

                    // Test decoding
                    if let Ok(decoded) = tokenizer.decode(&tokens.input_ids) {
                        // Decoded text shouldn't be extremely longer than original
                        prop_assert!(decoded.len() <= text.len() * 3);
                    }
                },
                Err(_) => {
                    // Some invalid inputs may fail, which is acceptable
                },
            }

            Ok(())
        }

        proptest!(|(text in ".*{0,1000}")| {
            tokenize_test(text)?;
        });
    }

    /// Fuzz test for pipeline configuration
    #[test]
    fn fuzz_pipeline_config() {
        fn config_test(
            _device: String,
            batch_size: usize,
            max_length: usize,
            _temperature: f32,
        ) -> std::result::Result<(), TestCaseError> {
            // Test pipeline configuration with random parameters
            use trustformers::pipeline::PipelineOptions;

            let config = PipelineOptions {
                device: None, // Skip device for now due to type issues
                batch_size: Some(batch_size),
                max_length: Some(max_length),
                ..Default::default()
            };

            // Configuration creation should not panic
            prop_assert!(config.batch_size.unwrap_or(1) > 0);
            prop_assert!(config.max_length.unwrap_or(1) > 0);
            prop_assert!(config.truncation || !config.truncation);

            Ok(())
        }

        proptest!(|(
            device in prop::sample::select(vec!["cpu", "cuda", "mps", "auto"]),
            batch_size in 1usize..=128,
            max_length in 1usize..=8192,
            temperature in 0.0f32..=2.0f32,
        )| {
            config_test(device.to_string(), batch_size, max_length, temperature)?;
        });
    }

    /// Fuzz test for model metadata parsing
    #[test]
    fn fuzz_model_metadata() {
        fn metadata_test(
            metadata: HashMap<String, String>,
        ) -> std::result::Result<(), TestCaseError> {
            // Test metadata handling with arbitrary key-value pairs
            let metadata_json = serde_json::to_string(&metadata);

            match metadata_json {
                Ok(json_str) => {
                    // Test JSON roundtrip
                    if let Ok(parsed_metadata) =
                        serde_json::from_str::<HashMap<String, String>>(&json_str)
                    {
                        prop_assert_eq!(metadata.len(), parsed_metadata.len());
                    }
                },
                Err(_) => {
                    // Some metadata might not be serializable, which is fine
                },
            }

            Ok(())
        }

        proptest!(|(
            metadata in prop::collection::hash_map(
                "[a-zA-Z0-9_-]{1,50}",    // Keys
                ".*{0,200}",              // Values
                0..20                     // Size
            )
        )| {
            metadata_test(metadata)?;
        });
    }

    /// Fuzz test for error handling and recovery
    #[test]
    fn fuzz_error_handling() {
        fn error_test(
            error_message: String,
            _error_code: u32,
        ) -> std::result::Result<(), TestCaseError> {
            // Test error creation and handling
            let error = trustformers::error::TrustformersError::Pipeline {
                message: error_message.clone(),
                pipeline_type: "test-pipeline".to_string(),
                suggestion: Some("test suggestion".to_string()),
                recovery_actions: vec![],
            };

            // Error formatting should not panic
            let error_string = format!("{}", error);
            prop_assert!(!error_string.is_empty());

            // Error debug formatting should not panic
            let debug_string = format!("{:?}", error);
            prop_assert!(!debug_string.is_empty());

            Ok(())
        }

        proptest!(|(
            error_message in ".*{0,500}",
            error_code in 0u32..=10000u32,
        )| {
            error_test(error_message, error_code)?;
        });
    }

    /// Fuzz test for memory operations
    #[test]
    fn fuzz_memory_operations() {
        fn memory_test(
            allocation_size: usize,
            alignment: usize,
        ) -> std::result::Result<(), TestCaseError> {
            // Skip unreasonable sizes
            if allocation_size > 1_000_000 || alignment == 0 || !alignment.is_power_of_two() {
                return Ok(());
            }

            // Test memory pool operations
            use trustformers::memory_pool::{MemoryPool, MemoryPoolConfig};

            let config = MemoryPoolConfig {
                initial_size: allocation_size * 2,
                max_size: allocation_size * 10,
                alignment: 8,
                ..Default::default()
            };

            if let Ok(pool) = MemoryPool::new(config) {
                // Test allocation
                if let Ok(memory) = pool.allocate(allocation_size) {
                    // Test deallocation
                    let _ = pool.deallocate(memory);
                }
            }

            Ok(())
        }

        proptest!(|(
            allocation_size in 1usize..=100_000,
            alignment_exp in 0u8..=10,
        )| {
            let alignment = 1usize << alignment_exp;
            memory_test(allocation_size, alignment)?;
        });
    }

    /// Fuzz test for hub operations
    #[test]
    fn fuzz_hub_operations() {
        fn hub_test(
            model_id: String,
            revision: Option<String>,
        ) -> std::result::Result<(), TestCaseError> {
            // Test hub model cache checking
            use trustformers::hub::is_cached;

            // Cache check should not panic
            let _is_cached = is_cached(&model_id, revision.as_deref()).unwrap_or(false);

            Ok(())
        }

        proptest!(|(
            model_id in "[a-zA-Z0-9/_-]{1,100}",
            revision in prop::option::of("[a-zA-Z0-9._-]{0,50}"),
        )| {
            hub_test(model_id, revision)?;
        });
    }

    /// Fuzz test for configuration validation
    #[test]
    fn fuzz_config_validation() {
        fn config_validation_test(
            learning_rate: f64,
            batch_size: i32,
            max_steps: i64,
        ) -> std::result::Result<(), TestCaseError> {
            use trustformers::config_management::ConfigValidator;

            // Create test configuration
            let mut config = HashMap::new();
            config.insert("learning_rate".to_string(), learning_rate.to_string());
            config.insert("batch_size".to_string(), batch_size.to_string());
            config.insert("max_steps".to_string(), max_steps.to_string());

            let validator = ConfigValidator::new();
            let config_value = serde_json::to_value(&config).expect("operation failed in test");
            let schema = trustformers::config_management::ConfigSchema {
                name: "test_schema".to_string(),
                version: "1.0".to_string(),
                description: "Test schema for fuzzing".to_string(),
                fields: HashMap::new(),
                required_fields: HashSet::new(),
                conditional_requirements: Vec::new(),
            };

            // Validation should not panic
            let _result = validator.validate(&config_value, &schema);

            Ok(())
        }

        proptest!(|(
            learning_rate in -1.0f64..=1.0f64,
            batch_size in -100i32..=1000i32,
            max_steps in -1000i64..=100000i64,
        )| {
            config_validation_test(learning_rate, batch_size, max_steps)?;
        });
    }

    /// Fuzz test for streaming operations
    #[test]
    fn fuzz_streaming_operations() {
        fn streaming_test(
            chunk_size: usize,
            buffer_size: usize,
            timeout_ms: u64,
        ) -> std::result::Result<(), TestCaseError> {
            use trustformers::pipeline::streaming::StreamConfig;

            // Skip unreasonable values
            if chunk_size == 0 || buffer_size == 0 || timeout_ms > 60000 {
                return Ok(());
            }

            // Test stream configuration creation
            let config = StreamConfig {
                buffer_size,
                max_concurrent: 4,
                backpressure_threshold: 0.8,
                timeout_ms,
                enable_partial_results: true,
                enable_transformations: true,
                batch_size: Some(chunk_size),
                flush_interval_ms: 100,
            };

            // Configuration should be valid
            prop_assert!(config.buffer_size > 0);
            prop_assert!(config.batch_size.unwrap_or(0) > 0);
            prop_assert!(
                config.backpressure_threshold >= 0.0 && config.backpressure_threshold <= 1.0
            );

            Ok(())
        }

        proptest!(|(
            chunk_size in 1usize..=10000,
            buffer_size in 1usize..=10000,
            timeout_ms in 1u64..=5000,
        )| {
            streaming_test(chunk_size, buffer_size, timeout_ms)?;
        });
    }
}

/// Regression fuzzing tests
/// Tests for previously discovered issues to prevent regressions
#[cfg(test)]
mod regression_fuzz_tests {
    use super::*;

    /// Test for empty input handling regression
    #[test]
    fn fuzz_empty_inputs() {
        // Test various components with empty inputs
        let empty_text = "";
        let empty_vec: Vec<f32> = vec![];
        let _empty_shape: Vec<usize> = vec![];

        // Tokenizer should handle empty text gracefully
        let vocab = HashMap::new();
        let merges = Vec::new();
        let bpe_tokenizer = trustformers::tokenizers::bpe::BPETokenizer::new(vocab, merges);
        let tokenizer = TokenizerWrapper::BPE(bpe_tokenizer);
        {
            let _result = tokenizer.encode(empty_text);
        }

        // Tensor operations should handle empty data gracefully
        let _tensor_result = Tensor::from_slice(&empty_vec, &[0]);

        // Configuration should handle empty values
        let empty_config: HashMap<String, String> = HashMap::new();
        let _json_result = serde_json::to_string(&empty_config);
    }

    /// Test for unicode handling regression
    #[test]
    fn fuzz_unicode_inputs() {
        let long_text = "a".repeat(10000);
        let unicode_texts = vec![
            "Hello, 世界!",
            "🦀 Rust is great! 🚀",
            "Здравствуй мир",
            "مرحبا بالعالم",
            "नमस्ते दुनिया",
            "こんにちは世界",
            "",
            "\n\t\r",
            &long_text,
        ];

        for text in unicode_texts {
            // Test tokenization with unicode
            let vocab = HashMap::new();
            let merges = Vec::new();
            let bpe_tokenizer = trustformers::tokenizers::bpe::BPETokenizer::new(vocab, merges);
            let tokenizer = TokenizerWrapper::BPE(bpe_tokenizer);
            let _result = tokenizer.encode(text);

            // Test error message creation with unicode
            let error = trustformers::error::TrustformersError::Pipeline {
                message: text.to_string(),
                pipeline_type: "test-pipeline".to_string(),
                suggestion: Some("test suggestion".to_string()),
                recovery_actions: vec![],
            };
            let _formatted = format!("{}", error);
        }
    }

    /// Test for numerical edge cases regression
    #[test]
    fn fuzz_numerical_edge_cases() {
        let edge_values = vec![
            f32::INFINITY,
            f32::NEG_INFINITY,
            f32::NAN,
            f32::MIN,
            f32::MAX,
            0.0,
            -0.0,
            f32::EPSILON,
            1.0,
            -1.0,
        ];

        for value in edge_values {
            // Test tensor creation with edge values
            let data = vec![value; 10];
            let _tensor_result = Tensor::from_slice(&data, &[2, 5]);

            // Test configuration with edge values
            let mut config = HashMap::new();
            config.insert("temperature".to_string(), value.to_string());
            let _json_result = serde_json::to_string(&config);
        }
    }
}

/// Performance fuzzing tests
/// Tests for performance regression under various load conditions
#[cfg(test)]
mod performance_fuzz_tests {
    use super::*;
    use std::time::{Duration, Instant};

    /// Test that operations complete within reasonable time bounds
    #[test]
    fn fuzz_performance_bounds() {
        proptest!(|(
            size in 1usize..=1000,
            iterations in 1usize..=100,
        )| {
            let start = Instant::now();

            // Perform operations that should be reasonably fast
            for _ in 0..iterations {
                let data: Vec<f32> = (0..size).map(|i| i as f32).collect();
                if let Ok(tensor) = Tensor::from_slice(&data, &[size]) {
                    let _ = tensor.data();
                    let _ = tensor.shape();
                }
            }

            let elapsed = start.elapsed();
            prop_assert!(elapsed < Duration::from_secs(10), "Operations took too long: {:?}", elapsed);
        });
    }

    /// Test memory usage stays within bounds
    #[test]
    fn fuzz_memory_bounds() {
        use std::alloc::{GlobalAlloc, Layout, System};

        // Test that memory allocations are reasonable
        proptest!(|(
            allocation_count in 1usize..=100,
            allocation_size in 1usize..=1000,
        )| {
            let mut allocations = Vec::new();

            for _ in 0..allocation_count {
                if let Ok(layout) = Layout::array::<u8>(allocation_size) {
                    // This is just a test - in real code, use proper memory management
                    let ptr = unsafe { System.alloc(layout) };
                    if !ptr.is_null() {
                        allocations.push((ptr, layout));
                    }
                }
            }

            // Clean up allocations
            for (ptr, layout) in allocations {
                unsafe { System.dealloc(ptr, layout) };
            }

            prop_assert!(true); // Test that we can allocate and deallocate without issues
        });
    }
}