Review of a software could range from an informal walk-through without comments to peer-reviews that capture comments, panel-reviews and verification-cycles.
This project is about analyzing the text in a series of reviews done over a period of time.
At a top level, the program runs in two processes - the sender and the receiver.
The Sender is responsible for reading the reviews from an input. The Receiver analyzes the data.
The Sender sends data to the Receiver using console redirection. Run them on the command line as follows:
sender-executable | receiver-executable
This would make the console-writes of the sender become the console-reads of the receiver.
The naming of source files within the Sender and within the Receiver give their internal decomposition.
This project follows the practices and tools listed here.
We document the interface between the Sender and the Receiver as test cases.
The Sender and Receiver are testable on their own:
- The Sender is testable without the Receiver - so we can develop for another data-source, test and be confident about integration.
- The Receiver is testable without the Sender - so we can enhance without re-testing against all Receivers again.
When we invoke the Sender executable with a CSV file on the command line, it outputs a series of words contained in that file.
When we give a column-filter to the Sender, it outputs words from that column alone.
The Receiver inputs the words from the console and performs a word-count. It writes the word-counts to a CSV file, which a spreadsheet application can open.
The Receiver removes stop words (words like a, the, and...) that do not contribute to our analysis.
When the input CSV contains the dates of the reviews, the Receiver can associate the word-counts with the date(s) on which a review used that word. This is to enable an analysis of how reviews have progressed over time.
See here for the evaluation criteria of this exercise.