// OCR.space: 80+ lines of infrastructure before business logic
public class OcrSpaceApiClient : IDisposable
{
    private readonly HttpClient _httpClient;
    private readonly string _apiKey;
    private readonly SemaphoreSlim _rateLimiter; // You implement this

    public OcrSpaceApiClient(string apiKey)
    {
        _httpClient = new HttpClient();
        _httpClient.Timeout = TimeSpan.FromSeconds(120);
        _rateLimiter = new SemaphoreSlim(60, 60); // Free tier: 60/min
    }
    // ... 70+ more lines of HTTP plumbing follow
}

// OCR.space: 80+ lines of infrastructure before business logic
public class OcrSpaceApiClient : IDisposable
{
    private readonly HttpClient _httpClient;
    private readonly string _apiKey;
    private readonly SemaphoreSlim _rateLimiter; // You implement this

    public OcrSpaceApiClient(string apiKey)
    {
        _httpClient = new HttpClient();
        _httpClient.Timeout = TimeSpan.FromSeconds(120);
        _rateLimiter = new SemaphoreSlim(60, 60); // Free tier: 60/min
    }
    // ... 70+ more lines of HTTP plumbing follow
}

// IronOCR: no client to build, no rate limiter to manage
IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";
var result = new IronTesseract().Read("document.jpg");
Console.WriteLine(result.Text);

// IronOCR: no client to build, no rate limiter to manage
IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";
var result = new IronTesseract().Read("document.jpg");
Console.WriteLine(result.Text);

// Before (OCR.space — manually written infrastructure)
using System.Net.Http;
using System.Text.Json;
using System.Threading;

// After (IronOCR)
using IronOcr;

// Before (OCR.space — manually written infrastructure)
using System.Net.Http;
using System.Text.Json;
using System.Threading;

// After (IronOCR)
using IronOcr;

// Program.cs or application startup
IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

// Program.cs or application startup
IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

// MultipartFormDataContent: file upload to cloud on every request
public async Task<string> UploadAndExtract(string imagePath)
{
    using var content = new MultipartFormDataContent();
    var imageBytes = File.ReadAllBytes(imagePath);

    // Document is transmitted to OCR.space servers here
    content.Add(new ByteArrayContent(imageBytes), "file", Path.GetFileName(imagePath));
    content.Add(new StringContent(_apiKey), "apikey");
    content.Add(new StringContent("eng"), "language");
    content.Add(new StringContent("2"), "OCREngine"); // Select Engine 2

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", content);
    response.EnsureSuccessStatusCode();

    string json = await response.Content.ReadAsStringAsync();
    using var doc = JsonDocument.Parse(json);

    // Navigate JSON tree manually — no typed result
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// MultipartFormDataContent: file upload to cloud on every request
public async Task<string> UploadAndExtract(string imagePath)
{
    using var content = new MultipartFormDataContent();
    var imageBytes = File.ReadAllBytes(imagePath);

    // Document is transmitted to OCR.space servers here
    content.Add(new ByteArrayContent(imageBytes), "file", Path.GetFileName(imagePath));
    content.Add(new StringContent(_apiKey), "apikey");
    content.Add(new StringContent("eng"), "language");
    content.Add(new StringContent("2"), "OCREngine"); // Select Engine 2

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", content);
    response.EnsureSuccessStatusCode();

    string json = await response.Content.ReadAsStringAsync();
    using var doc = JsonDocument.Parse(json);

    // Navigate JSON tree manually — no typed result
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// OcrInput replaces the entire upload + JSON pipeline
public string ExtractFromFile(string imagePath)
{
    using var input = new OcrInput();
    input.LoadImage(imagePath); // Stays local — no network call

    var ocr = new IronTesseract();
    var result = ocr.Read(input);

    return result.Text; // Typed property — no JSON navigation
}

// OcrInput replaces the entire upload + JSON pipeline
public string ExtractFromFile(string imagePath)
{
    using var input = new OcrInput();
    input.LoadImage(imagePath); // Stays local — no network call

    var ocr = new IronTesseract();
    var result = ocr.Read(input);

    return result.Text; // Typed property — no JSON navigation
}

// base64Image parameter: read → encode → embed in form → POST → parse
public async Task<string> ExtractViaBase64(string imagePath)
{
    byte[] imageBytes = await File.ReadAllBytesAsync(imagePath);
    string base64Image = Convert.ToBase64String(imageBytes); // Mandatory encoding step

    // Embed as data URI — adds 33% overhead to payload size
    string mimeType = "image/png";
    string dataUri = $"data:{mimeType};base64,{base64Image}";

    var formContent = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", dataUri),
        new KeyValuePair<string, string>("language", "eng"),
        new KeyValuePair<string, string>("isOverlayRequired", "false")
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", formContent);
    string json = await response.Content.ReadAsStringAsync();

    using var doc = JsonDocument.Parse(json);
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// base64Image parameter: read → encode → embed in form → POST → parse
public async Task<string> ExtractViaBase64(string imagePath)
{
    byte[] imageBytes = await File.ReadAllBytesAsync(imagePath);
    string base64Image = Convert.ToBase64String(imageBytes); // Mandatory encoding step

    // Embed as data URI — adds 33% overhead to payload size
    string mimeType = "image/png";
    string dataUri = $"data:{mimeType};base64,{base64Image}";

    var formContent = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", dataUri),
        new KeyValuePair<string, string>("language", "eng"),
        new KeyValuePair<string, string>("isOverlayRequired", "false")
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", formContent);
    string json = await response.Content.ReadAsStringAsync();

    using var doc = JsonDocument.Parse(json);
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// LoadImage(bytes): raw bytes accepted directly — no encoding
public string ExtractFromBytes(byte[] imageBytes)
{
    using var input = new OcrInput();
    input.LoadImage(imageBytes); //NeinBase64, no data URI, no overhead

    var result = new IronTesseract().Read(input);
    return result.Text;
}

// LoadImage(bytes): raw bytes accepted directly — no encoding
public string ExtractFromBytes(byte[] imageBytes)
{
    using var input = new OcrInput();
    input.LoadImage(imageBytes); //NeinBase64, no data URI, no overhead

    var result = new IronTesseract().Read(input);
    return result.Text;
}

// OCREngine parameter: binary choice, no control over why accuracy differs
public async Task<string> ExtractWithEngineSelection(
    string imagePath,
    bool useHighAccuracyEngine = true)
{
    byte[] imageBytes = await File.ReadAllBytesAsync(imagePath);
    string base64 = Convert.ToBase64String(imageBytes);

    var formContent = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", $"data:image/png;base64,{base64}"),
        new KeyValuePair<string, string>("language", "eng"),
        // Engine 1 = faster, Engine 2 = higher accuracy — binary choice only
        new KeyValuePair<string, string>("OCREngine", useHighAccuracyEngine ? "2" : "1"),
        new KeyValuePair<string, string>("scale", "true"),
        new KeyValuePair<string, string>("detectOrientation", "true")
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", formContent);
    string json = await response.Content.ReadAsStringAsync();

    using var doc = JsonDocument.Parse(json);
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// OCREngine parameter: binary choice, no control over why accuracy differs
public async Task<string> ExtractWithEngineSelection(
    string imagePath,
    bool useHighAccuracyEngine = true)
{
    byte[] imageBytes = await File.ReadAllBytesAsync(imagePath);
    string base64 = Convert.ToBase64String(imageBytes);

    var formContent = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", $"data:image/png;base64,{base64}"),
        new KeyValuePair<string, string>("language", "eng"),
        // Engine 1 = faster, Engine 2 = higher accuracy — binary choice only
        new KeyValuePair<string, string>("OCREngine", useHighAccuracyEngine ? "2" : "1"),
        new KeyValuePair<string, string>("scale", "true"),
        new KeyValuePair<string, string>("detectOrientation", "true")
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", formContent);
    string json = await response.Content.ReadAsStringAsync();

    using var doc = JsonDocument.Parse(json);
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// Preprocessing pipeline: fix the document, not the engine selection
public string ExtractWithPreprocessing(string imagePath)
{
    using var input = new OcrInput();
    input.LoadImage(imagePath);

    // Apply filters that match the document's specific quality issues
    input.Deskew();         // Correct rotation — replaces detectOrientation
    input.DeNoise();        // Remove noise from fax/photocopier artifacts
    input.Contrast();       // Enhance contrast on low-quality scans
    input.Scale(200);       // Upscale small or low-DPI images

    var ocr = new IronTesseract();
    var result = ocr.Read(input);

    Console.WriteLine($"Confidence: {result.Confidence}%"); //Neinequivalent in OCR.space
    return result.Text;
}

// Preprocessing pipeline: fix the document, not the engine selection
public string ExtractWithPreprocessing(string imagePath)
{
    using var input = new OcrInput();
    input.LoadImage(imagePath);

    // Apply filters that match the document's specific quality issues
    input.Deskew();         // Correct rotation — replaces detectOrientation
    input.DeNoise();        // Remove noise from fax/photocopier artifacts
    input.Contrast();       // Enhance contrast on low-quality scans
    input.Scale(200);       // Upscale small or low-DPI images

    var ocr = new IronTesseract();
    var result = ocr.Read(input);

    Console.WriteLine($"Confidence: {result.Confidence}%"); //Neinequivalent in OCR.space
    return result.Text;
}

// OCR.space: one language per call — multi-language requires multiple requests
public async Task<string> ExtractMultiLanguage(string imagePath)
{
    // First pass: English
    string englishText = await ExtractWithLanguage(imagePath, "eng");

    // Second pass: French (consumes another rate-limit slot, another API call)
    string frenchText = await ExtractWithLanguage(imagePath, "fre");

    // Manually merge results — no way to know which text belongs to which language
    return $"{englishText}\n{frenchText}";
}

private async Task<string> ExtractWithLanguage(string imagePath, string langCode)
{
    byte[] imageBytes = await File.ReadAllBytesAsync(imagePath);
    string base64 = Convert.ToBase64String(imageBytes);

    var content = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", $"data:image/png;base64,{base64}"),
        new KeyValuePair<string, string>("language", langCode) // One language per call
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", content);
    string json = await response.Content.ReadAsStringAsync();

    using var doc = JsonDocument.Parse(json);
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// OCR.space: one language per call — multi-language requires multiple requests
public async Task<string> ExtractMultiLanguage(string imagePath)
{
    // First pass: English
    string englishText = await ExtractWithLanguage(imagePath, "eng");

    // Second pass: French (consumes another rate-limit slot, another API call)
    string frenchText = await ExtractWithLanguage(imagePath, "fre");

    // Manually merge results — no way to know which text belongs to which language
    return $"{englishText}\n{frenchText}";
}

private async Task<string> ExtractWithLanguage(string imagePath, string langCode)
{
    byte[] imageBytes = await File.ReadAllBytesAsync(imagePath);
    string base64 = Convert.ToBase64String(imageBytes);

    var content = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", $"data:image/png;base64,{base64}"),
        new KeyValuePair<string, string>("language", langCode) // One language per call
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", content);
    string json = await response.Content.ReadAsStringAsync();

    using var doc = JsonDocument.Parse(json);
    return doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("ParsedText")
        .GetString() ?? string.Empty;
}

// IronOCR: multiple languages in a single read — one pass, correct output
public string ExtractMultiLanguage(string imagePath)
{
    var ocr = new IronTesseract();

    // Combine languages with + operator — processed simultaneously
    ocr.Language = OcrLanguage.English + OcrLanguage.French + OcrLanguage.German;

    var result = ocr.Read(imagePath);
    return result.Text; // Correctly interleaved multilingual output
}

// IronOCR: multiple languages in a single read — one pass, correct output
public string ExtractMultiLanguage(string imagePath)
{
    var ocr = new IronTesseract();

    // Combine languages with + operator — processed simultaneously
    ocr.Language = OcrLanguage.English + OcrLanguage.French + OcrLanguage.German;

    var result = ocr.Read(imagePath);
    return result.Text; // Correctly interleaved multilingual output
}

// OCR.space: plain text only — no structure, no coordinates
public async Task<string> ExtractInvoiceFields(string invoicePath)
{
    byte[] invoiceBytes = await File.ReadAllBytesAsync(invoicePath);
    string base64 = Convert.ToBase64String(invoiceBytes);

    var content = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", $"data:application/pdf;base64,{base64}"),
        new KeyValuePair<string, string>("filetype", "PDF"),
        new KeyValuePair<string, string>("language", "eng"),
        // isOverlayRequired=true returns word boxes, but only as raw JSON coordinates
        new KeyValuePair<string, string>("isOverlayRequired", "true")
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", content);
    string json = await response.Content.ReadAsStringAsync();

    // Navigate deeply-nested JSON to find word boxes — no typed models
    using var doc = JsonDocument.Parse(json);
    var overlay = doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("TextOverlay");

    // Parse word coordinate arrays manually — fragile JSON path traversal
    var wordData = new List<(string word, int x, int y)>();
    foreach (var line in overlay.GetProperty("Lines").EnumerateArray())
    {
        foreach (var word in line.GetProperty("Words").EnumerateArray())
        {
            string wordText = word.GetProperty("WordText").GetString() ?? "";
            int left = word.GetProperty("Left").GetInt32();
            int top = word.GetProperty("Top").GetInt32();
            wordData.Add((wordText, left, top));
        }
    }

    // Reconstruct full text from raw JSON — still no typed result
    return string.Join(" ", wordData.Select(w => w.word));
}

// OCR.space: plain text only — no structure, no coordinates
public async Task<string> ExtractInvoiceFields(string invoicePath)
{
    byte[] invoiceBytes = await File.ReadAllBytesAsync(invoicePath);
    string base64 = Convert.ToBase64String(invoiceBytes);

    var content = new FormUrlEncodedContent(new[]
    {
        new KeyValuePair<string, string>("apikey", _apiKey),
        new KeyValuePair<string, string>("base64Image", $"data:application/pdf;base64,{base64}"),
        new KeyValuePair<string, string>("filetype", "PDF"),
        new KeyValuePair<string, string>("language", "eng"),
        // isOverlayRequired=true returns word boxes, but only as raw JSON coordinates
        new KeyValuePair<string, string>("isOverlayRequired", "true")
    });

    var response = await _httpClient.PostAsync("https://api.ocr.space/parse/image", content);
    string json = await response.Content.ReadAsStringAsync();

    // Navigate deeply-nested JSON to find word boxes — no typed models
    using var doc = JsonDocument.Parse(json);
    var overlay = doc.RootElement
        .GetProperty("ParsedResults")[0]
        .GetProperty("TextOverlay");

    // Parse word coordinate arrays manually — fragile JSON path traversal
    var wordData = new List<(string word, int x, int y)>();
    foreach (var line in overlay.GetProperty("Lines").EnumerateArray())
    {
        foreach (var word in line.GetProperty("Words").EnumerateArray())
        {
            string wordText = word.GetProperty("WordText").GetString() ?? "";
            int left = word.GetProperty("Left").GetInt32();
            int top = word.GetProperty("Top").GetInt32();
            wordData.Add((wordText, left, top));
        }
    }

    // Reconstruct full text from raw JSON — still no typed result
    return string.Join(" ", wordData.Select(w => w.word));
}

// IronOCR: full document hierarchy — typed, no JSON, no coordinates parsing
public void ExtractInvoiceFields(string invoicePath)
{
    var ocr = new IronTesseract();
    var result = ocr.Read(invoicePath);

    // Access the full document hierarchy — all strongly typed
    foreach (var page in result.Pages)
    {
        foreach (var paragraph in page.Paragraphs)
        {
            Console.WriteLine($"Paragraph at ({paragraph.X}, {paragraph.Y}): {paragraph.Text}");
        }

        foreach (var word in page.Words)
        {
            // Word-level confidence — identify low-quality extractions
            if (word.Confidence < 70)
                Console.WriteLine($"Low confidence word '{word.Text}' at ({word.X}, {word.Y})");
        }
    }

    // Or use region-based OCR to target specific invoice zones directly
    var totalRegion = new CropRectangle(400, 700, 200, 50); // Bottom-right total field
    using var input = new OcrInput();
    input.LoadImage(invoicePath, totalRegion);
    string totalText = ocr.Read(input).Text;
    Console.WriteLine($"Invoice total: {totalText}");
}

// IronOCR: full document hierarchy — typed, no JSON, no coordinates parsing
public void ExtractInvoiceFields(string invoicePath)
{
    var ocr = new IronTesseract();
    var result = ocr.Read(invoicePath);

    // Access the full document hierarchy — all strongly typed
    foreach (var page in result.Pages)
    {
        foreach (var paragraph in page.Paragraphs)
        {
            Console.WriteLine($"Paragraph at ({paragraph.X}, {paragraph.Y}): {paragraph.Text}");
        }

        foreach (var word in page.Words)
        {
            // Word-level confidence — identify low-quality extractions
            if (word.Confidence < 70)
                Console.WriteLine($"Low confidence word '{word.Text}' at ({word.X}, {word.Y})");
        }
    }

    // Or use region-based OCR to target specific invoice zones directly
    var totalRegion = new CropRectangle(400, 700, 200, 50); // Bottom-right total field
    using var input = new OcrInput();
    input.LoadImage(invoicePath, totalRegion);
    string totalText = ocr.Read(input).Text;
    Console.WriteLine($"Invoice total: {totalText}");
}

// Before: async because OCR.space required network I/O
public async Task<string> ProcessDocumentAsync(string path)
{
    return await _ocrSpaceClient.ExtractTextAsync(path); // Network wait
}

// After: synchronous — no network, no async needed
public string ProcessDocument(string path)
{
    return _ocr.Read(path).Text; // Local execution
}

// Before: async because OCR.space required network I/O
public async Task<string> ProcessDocumentAsync(string path)
{
    return await _ocrSpaceClient.ExtractTextAsync(path); // Network wait
}

// After: synchronous — no network, no async needed
public string ProcessDocument(string path)
{
    return _ocr.Read(path).Text; // Local execution
}

// Startup.cs or Program.cs — once only
IronOcr.License.LicenseKey = Environment.GetEnvironmentVariable("IRONOCR_LICENSE_KEY");

// Startup.cs or Program.cs — once only
IronOcr.License.LicenseKey = Environment.GetEnvironmentVariable("IRONOCR_LICENSE_KEY");

// OCR.space: pre-validation to avoid wasting quota on large files
var fileInfo = new FileInfo(filePath);
if (fileInfo.Length > 5 * 1024 * 1024)
    throw new InvalidOperationException("File exceeds 5MB free tier limit.");

// OCR.space: pre-validation to avoid wasting quota on large files
var fileInfo = new FileInfo(filePath);
if (fileInfo.Length > 5 * 1024 * 1024)
    throw new InvalidOperationException("File exceeds 5MB free tier limit.");

// IronOCR: typed result — no JSON path fragility
var result = new IronTesseract().Read("document.jpg");

if (result.Confidence < 50)
    Console.WriteLine("Low confidence — consider preprocessing");
else
    Console.WriteLine(result.Text);

// IronOCR: typed result — no JSON path fragility
var result = new IronTesseract().Read("document.jpg");

if (result.Confidence < 50)
    Console.WriteLine("Low confidence — consider preprocessing");
else
    Console.WriteLine(result.Text);

# Find all OCR.space HTTP calls
grep -rn "ocr.space" --include="*.cs" .

# Find all base64 encoding related to OCR
grep -rn "base64Image\|Convert.ToBase64String" --include="*.cs" .

# Find rate limiter and retry logic
grep -rn "SemaphoreSlim\|TooManyRequests\|exponential" --include="*.cs" .

# Find JSON parsing for OCR responses
grep -rn "ParsedResults\|IsErroredOnProcessing\|FileParseExitCode" --include="*.cs" .

# Find custom OCR models and exception types
grep -rn "OcrSpaceException\|OcrResult\b" --include="*.cs" .

# Find API key configuration
grep -rn "apikey\|OcrSpaceApiKey\|ocr_space" --include="*.cs" --include="*.json" .

# Find all OCR.space HTTP calls
grep -rn "ocr.space" --include="*.cs" .

# Find all base64 encoding related to OCR
grep -rn "base64Image\|Convert.ToBase64String" --include="*.cs" .

# Find rate limiter and retry logic
grep -rn "SemaphoreSlim\|TooManyRequests\|exponential" --include="*.cs" .

# Find JSON parsing for OCR responses
grep -rn "ParsedResults\|IsErroredOnProcessing\|FileParseExitCode" --include="*.cs" .

# Find custom OCR models and exception types
grep -rn "OcrSpaceException\|OcrResult\b" --include="*.cs" .

# Find API key configuration
grep -rn "apikey\|OcrSpaceApiKey\|ocr_space" --include="*.cs" --include="*.json" .