The current implementation takes the average execution time of each item to determine the estimated time remaining. Depending on the scenario, item processing may go slower for each item as the processing progresses, such as a large database migration that has more and more data to query, processing images in order of their file size, etc.

Let's say there are 100 items to process and processing time increases as they are processed. The first twenty items take one second each, the next twenty take five seconds each, and the next twenty take ten seconds each. Using the current implementation, that's an average of 5.3 seconds per item.

Rather than using an overall average, I'd like to implement a parameter in the constructor that takes a sample mode and a sample size. If we pass in a calculation mode of count and a sample size of 40, the code would look at the last forty items processed to get an average time of 7.5 seconds per item. This time would be used to calculate the estimated time remaining for the rest of the items to be processed. I would also like to allow for a calculation mode of percent to look at the last X percent of items to get the average.

Examples:

// Use the last 40 items to determine an average.
$estimator = new \PHPEstimator\Estimator(
   100,
   ['calculation_mode' => 'count', 'sample_size' => 40]
);

// Use the last 10 percent of the total items to determine an average.
$estimator = new \PHPEstimator\Estimator(
   100,
   ['calculation_mode' => 'percent', 'sample_size' => 10]
);

The code would not switch over to using a different calculation mode until the number of items processed reaches the sample size.