protected override Task <string> ExtractName(IDocument inputData, ExtractorSettings parserSettings)
{
var nameElement = inputData.QuerySelectorAll(parserSettings.Name).FirstOrDefault();
logger.LogInformation($"The processed part of the document by product name: {nameElement?.OuterHtml}");
return(Task.FromResult(nameElement?.TextContent.Trim()));
}
protected override bool IsCaughtByCaptcha(IDocument inputData, ExtractorSettings parserSettings)
{
if (inputData.Title == "Ой!")
{
logger.LogError($"Попали на капчу {inputData.Source.Text}");
return(true);
}
return(false);
}
public async Task <ProductData> Extract(T inputData, ExtractorSettings parserSettings)
{
if (IsCaughtByCaptcha(inputData, parserSettings))
{
throw new AggregateException($"Caught by CAPTCHA");
}
var name = await ExtractName(inputData, parserSettings);
var(price, discountPrice) = await ExtractPrice(inputData, parserSettings);
var additionalInformation = await ExtractAdditionalInformation(inputData, parserSettings);
if (discountPrice == null && price == null)
{
additionalInformation = await ExtractOutofstockInformation(inputData, parserSettings);
if (additionalInformation is null)
{
throw new FormatException("Unknown error while extract product data");
}
else
{
logger.LogInformation($"The item may be out of stock or has been removed. Info: {additionalInformation}");
}
}
return(new ProductData
{
Name = name,
Price = price,
DiscountPrice = discountPrice,
AdditionalInformation = additionalInformation,
Date = DateTime.Now
});
}
protected override Task <string> ExtractAdditionalInformation(IDocument htmlDocument, ExtractorSettings parserSettings)
{
if (parserSettings.AdditionalInformation == null)
{
return(Task.FromResult <string>(null));
}
var additionaInformation = new Dictionary <string, string>();
foreach (var keyValue in parserSettings.AdditionalInformation)
{
var element = htmlDocument.QuerySelectorAll(keyValue.Value).FirstOrDefault();
if (element == null)
{
logger.LogWarning($"Не удалось извлечь информацию о {keyValue.Key} по пути {keyValue.Value}");
}
var textContent = element?.TextContent ?? String.Empty;
additionaInformation.Add(keyValue.Key, TransformAdditionalInformation(textContent));
}
var options = new JsonSerializerOptions()
{
Encoder = JavaScriptEncoder.UnsafeRelaxedJsonEscaping
};
var additionalInformationString = JsonSerializer.Serialize(additionaInformation, options);
logger.LogInformation($"Найденная дополнительная информация {additionalInformationString}");
return(Task.FromResult(additionalInformationString));
}
protected override Task <(decimal?price, decimal?discountPrice)> ExtractPrice(IDocument inputData, ExtractorSettings parserSettings)
{
var discountPriceElement = inputData.QuerySelectorAll(parserSettings.DiscountHtmlPath).FirstOrDefault();
logger.LogInformation($"The processed part of the document by discount: {discountPriceElement?.OuterHtml}");
var discountPrice = discountPriceElement?.TextContent;
string price = default;
foreach (var priceHtmlPath in parserSettings.PriceHtmlPath)
{
var priceElement = inputData.QuerySelectorAll(priceHtmlPath).FirstOrDefault();
logger.LogInformation($"The processed part of the document by price {priceElement?.OuterHtml}");
price = priceElement?.TextContent;
if (!string.IsNullOrEmpty(price))
{
break;
}
}
if (discountPrice == null && price == null)
{
return(Task.FromResult <(decimal?price, decimal?discountPrice)>((null, null)));
}
if (discountPrice != null && price == null)
{
price = discountPrice;
discountPrice = null;
}
if (discountPrice != null)
{
discountPrice = TransformPrice(discountPrice);
discountPrice = ExtractPrice(discountPrice);
}
if (price != null)
{
price = TransformPrice(price);
price = ExtractPrice(price);
}
if (!decimal.TryParse(price, NumberStyles.AllowDecimalPoint, CultureInfo.InvariantCulture, out decimal priceValue))
{
throw new InvalidCastException($"Can not convert {nameof(price)}={price} to {typeof(decimal)}");
}
if (!decimal.TryParse(discountPrice, NumberStyles.AllowDecimalPoint, CultureInfo.InvariantCulture, out decimal discountPriceTemp) && discountPrice != null)
{
throw new InvalidCastException($"Can not convert {nameof(discountPrice)}={discountPrice} to {typeof(decimal)}");
}
decimal?discountPriceValue = discountPrice == null ? null : (decimal?)discountPriceTemp;
return(Task.FromResult(((decimal?)priceValue, discountPriceValue)));
}
protected override Task <string> ExtractOutofstockInformation(IDocument inputData, ExtractorSettings parserSettings) => Task.FromResult(inputData.QuerySelectorAll(parserSettings.OutOfStockHtmlPath).FirstOrDefault()?.TextContent);
protected override Task <(decimal?price, decimal?discountPrice)> ExtractPrice(string inputData, ExtractorSettings parserSettings)
{
//TODO Сделать вырез цены из скрипшота
ModelInput sampleData = new ModelInput {
ImageSource = inputData
};
// Make a single prediction on the sample data and print results
var prediction = predictionEnginePool.Predict(modelName: "CVPriceDetectionModel", example: sampleData);
var priceBox = prediction.BoundingBoxes?.Where(p => p.Label == "price").OrderByDescending(p => p.Score).First();
using Bitmap source = new Bitmap(sampleData.ImageSource);
using Bitmap resizeImage = ResizeBitmap(source, 800, 600);
Rectangle section = new Rectangle(new Point((int)priceBox.Left, (int)priceBox.Top), new Size((int)priceBox.Right - (int)priceBox.Left, (int)priceBox.Bottom - (int)priceBox.Top));
using Bitmap priceImage = CropImage(resizeImage, section);
var priceImagePath = Path.Combine(configuration.GetValue <string>("ImagesFolder"), $"{Path.GetFileNameWithoutExtension(inputData)}-price.png");
priceImage.Save(priceImagePath, System.Drawing.Imaging.ImageFormat.Png);
var root = configuration.GetValue <string>(WebHostDefaults.ContentRootKey);
using var tesseractEngine = new TesseractEngine($"{root}/CV/Tesseract", "eng+rus", EngineMode.Default);
using var image = Pix.LoadFromFile(priceImagePath);
using var page = tesseractEngine.Process(image);
var price = page.GetText();
File.Delete(priceImagePath);
if (price is null)
{
throw new AggregateException($"Can not find content from image {inputData}");
}
price = TransformPrice(price);
price = ExtractPrice(price);
if (!decimal.TryParse(price, NumberStyles.AllowDecimalPoint, CultureInfo.InvariantCulture, out decimal priceValue))
{
throw new InvalidCastException($"Can not convert {nameof(price)}={price} to {typeof(decimal)}");
}
return(Task.FromResult(((decimal?)priceValue, (decimal?)null)));
}
protected override Task <string> ExtractOutofstockInformation(string inputData, ExtractorSettings parserSettings) => Task.FromResult <string>(null);
protected override Task <string> ExtractName(string inputData, ExtractorSettings parserSettings) => Task.FromResult <string>(null);
protected override Task <string> ExtractAdditionalInformation(IDocument inputData, ExtractorSettings parserSettings) => Task.FromResult <string>(null);
protected override Task <(decimal?price, decimal?discountPrice)> ExtractPrice(IDocument inputData, ExtractorSettings parserSettings)
{
string priceHtmlElement = null;
string discountPriceHtmlElement = null;
int count = 0;
foreach (var priceData in GetPriceData(inputData))
{
var pricePrediction = predictionEnginePool.Predict(modelName: "MLPriceDetectionModel", example: priceData);
if (!bool.TryParse(pricePrediction.Prediction, out bool isPrice))
{
throw new InvalidCastException($"Can not convert {pricePrediction.Prediction} to {typeof(bool)}");
}
if (isPrice)
{
if (priceHtmlElement != null && count < MaxPriceElementsInterval)
{
discountPriceHtmlElement = priceData.HtmlElement;
logger.LogInformation($"Detect {nameof(discountPriceHtmlElement)}={discountPriceHtmlElement} with {pricePrediction.Score}=[{string.Join(';', pricePrediction.Score)}]");
}
if (priceHtmlElement == null)
{
priceHtmlElement = priceData.HtmlElement;
logger.LogInformation($"Detect {nameof(priceHtmlElement)}={priceHtmlElement} with {pricePrediction.Score}=[{string.Join(';', pricePrediction.Score)}]");
}
count = 0;
}
count++;
}
if (priceHtmlElement is null && discountPriceHtmlElement is null)
{
return(Task.FromResult(((decimal?)null, (decimal?)null)));
}
string price = ExtractPrice(priceHtmlElement);
string discountPrice = ExtractPrice(discountPriceHtmlElement);
if (!decimal.TryParse(price, NumberStyles.AllowDecimalPoint, CultureInfo.InvariantCulture, out decimal priceValue))
{
throw new InvalidCastException($"Can not convert {nameof(price)}={price} to {typeof(decimal)}");
}
if (!decimal.TryParse(discountPrice, NumberStyles.AllowDecimalPoint, CultureInfo.InvariantCulture, out decimal discountPriceTemp) && discountPrice != null)
{
throw new InvalidCastException($"Can not convert {nameof(discountPrice)}={discountPrice} to {typeof(decimal)}");
}
decimal?discountPriceValue = null;
if (discountPrice != null && discountPriceTemp < priceValue)
{
decimal temp = priceValue;
priceValue = (decimal)discountPriceValue;
discountPriceValue = temp;
}
return(Task.FromResult(((decimal?)priceValue, discountPriceValue)));
}
protected virtual bool IsCaughtByCaptcha(T inputData, ExtractorSettings parserSettings) => false;
protected abstract Task <string> ExtractOutofstockInformation(T inputData, ExtractorSettings parserSettings);
protected abstract Task <string> ExtractAdditionalInformation(T inputData, ExtractorSettings parserSettings);
protected abstract Task <(decimal?price, decimal?discountPrice)> ExtractPrice(T inputData, ExtractorSettings parserSettings);
protected abstract Task <string> ExtractName(T inputData, ExtractorSettings parserSettings);