Postgres Pages
July 6th, 2022
There's a great feeling of that "Aha!" moment when you peel away a layer of magic from a technology. I remember, back when I first began working with databases, I viewed the tabular results of a query as an end-all-be-all. That's a database table, those are the results, and that was it. To be fair to new engineers, there is a lot to take in, so it's not unreasonable to coast on some broad, nebulous understandings for a period. Here I'd like to examine a small piece of Postgres, the Page. Hopefully, instead of the magic of the unknown, the Page can become the magic of, "hey, that's pretty cool."
So, what is a page? Consider a couple of "big ideas" in Postgres - the table and the index. These two constructs are actually quite similar in that they're made of pages, they're arrays of pages. And the page itself is where items (think "rows") of data live. Your individual rows of data live in these pages. Consider a new table, first it begins with an array consisting of a single page. And as you add data, data which cannot fit within the current page, a new page is added to the array. And it keeps growing from there. So, now, you can imagine the fundamental idea of storing data within Postgres. You have a "table" made up of an array of pages and each page contains within it the actual data you insert into your table.
There's a little bit more to explore to pages. There is the PageHeaderData, ItemIdData, Items (rows), and Special space. There is, also, the "free space" within the page where additional data may be inserted. Understanding these isn't terribly critical. The big win is knowing that pages exist, knowing that tables are composed of them, and knowing that when you manipulate data (insert/update/delete) you're making changes to the items within the page and potentially adding a new page. Having a passing familiarity with the page will especially help when reasoning about things like index data access.
With that, I hope some of the fundamentals of Postgres data are a little less mystical. Lastly, Postgres has amazing documentation which is well worth the read for further understanding.