Processing that takes whatever portion of the input it can handle cheaply and reads a sample of the rest, rather than processing the whole thing.

Named after the algorithmic technique, so that a decision about how much of your document to actually read can be filed as an implementation detail.

• Sample Presented as Whole: The output covers the document, and the reading behind it covered part of it.
• Greedy Means Cheap: The word describes a strategy taking what is immediately available, not a system being generous.
• House-Cleaning Reality Check: Like reading the first and last page of a report on the train and telling the meeting you have been through it.

Supermarket Sweep (1990s) where contestants grab whatever is nearest and highest value in the time allowed, and entire aisles are never visited at all.

Reading part of the input and sampling the rest instead of processing all of it.

“Our ingestion layer applies greedy parsing and sub-sampling for optimal throughput.”

“The tool read parts of the file and skipped the rest, and the summary does not say which parts it skipped.”