Prepares PDFs for viewing on Overview

This processing step ensures a PDF has text and a thumbnail. That way, Overview can present it.

If the input JSON's wantSplitByPage is true, this step will output one (PDF+text+thumb) document per page and set metadata.pageNumber:N, where N is 1 for the first page.

If metadata is missing any of these tags, and the PDF can supply that data, the metadata value will be filled in the output (or each page of output):

• Title
• Author
• Subject
• Keywords
• Creator (program that created the document)
• Creation Date
• Modification Date

We use PDFium because it's fast and it starts up quickly. (Our framework starts one process per input file, so an out-of-memory error from one input file doesn't impact the next.)

Throughput goals: on a reasonable machine, for a small PDF...:

• Extracting text and thumbnail from a PDF should take <0.1s
• Generating a PDF per page (with text and thumbnail) should take <0.2s

1. Install Docker-CE.
2. Run docker build .

(Useful builds: docker build --target=test . will compile binaries and run unit tests. docker build --target=production . will produce a minimal image.)