Discussion
Studying Papyrix's support for FictionBook2 ebooks...
To open such ebooks Papyrix closes the existing file if any, creates a Fb2 object & invokes its load method, clears metadata setting its `type` property, extracts the ebook's title into the metadata followed by the author as well as filepaths to the cache & cover bitmap, & sets some initial pagecount properties.
The close method defers both to the metadata & that Fb2 object. To get the page count it (again) estimates 2KiB per page.
1/?
Its load method verifies the file exists, attempts to deserialize cached metadata via a helper & if successful immediately precedes to split the ebook into cached sections, opens the file or errors out long enough to retrieve its filesize, clears the author property, reopens the file via a helper detecting the text encoding & feeding it through Expat with callbacks, does similar to validate extracted section offsets, extracts (via a helper) the hierarchy from those nested sections, ...
3/?
... serializes cached metadata via a helper, splits the file up into cached individual sections via another helper, reloads table of contents from the cached metadata, sets a flag, & logs what it has done.
There's a simplified variant which doesn't process the sections.
Upon a start element it increments a counter validating its less than 100, skips anything inside <binary> tags, strips namespace prefixes, checks for binary tags & updates state to match, & ...
4/6?
... & updates flags for different tags occasionally extracting attributes.
Upon a FB2 close-tag Papyrix strips off the namespace & branches over the tag name to set appropriate flags or other properties. Possibly tidying them up or exiting early (if we only want metadata) from the parser.
Upon character data it appends to the appropriate property for the current flags.
I've already basically described the helper method which feeds the file through Expat with those callbacks.
5/6?
There's methods which defer to the Fb2 object's, if any, getters. Possibly reformatting results into common datastructures.
And there's a method which concatenates together the filepath to a cached section.
That Fb2 class is defined in a support library. Its constructor concatenates together the path to the file's cache & extracts the basename as a title. Its destructor finalizes its Expat XML parser if any.
Then there's that load method...
2/?