Switch from InvariantCultureIgnoreCase to OrdinalIgnoreCase when
checking file paths in archives for substrings indicating the file can
be ignored for performance gains.
Co-Authored-By: Dan Balasescu <smoogipoo@smgi.me>
Add a filename ignore list to ZipArchiveReader to filter out superfluous
OS-generated files from archives during the import process. In addition
to decreasing the size of files imported this allows imports of some
incorrectly-constructed archives. An example is the case of having
a __MACOSX directory next to a single directory with the actual files -
filtering out the former at ZipArchiveReader allows the fallback added
in #6170 to work.
After the preparatory introduction of LineBufferedReader, it is now
possible to introduce registration of fallback decoders that won't drop
input supplied in the first line of the file.
A fallback decoder is used when the magic in the first line of the file
does not match any of the other known decoders. In such a case,
the fallback decoder is constructed and provided a LineBufferedReader
instance. The process of matching magic only peeks the first non-empty
line, so it is available for re-reading in Decode() using ReadLine().
There can be only one fallback decoder per type; a second attempt of
registering a fallback will result in an exception to avoid bugs.
To address the issue of parsing failing on badly or non-headered files,
set the legacy decoders for Beatmaps and Storyboards as the fallbacks.
Due to non-trivial logic, several new, passing unit tests with possible
edge cases also included.
Add a line-buffered reader decorator operating on StreamReader
instances. The decorator has two main operations - PeekLine(), which
allows to see the next line in the stream without consuming it,
ReadLine(), which consumes and returns the next line in the stream, and
ReadToEnd() which reads all the remaining text in the stream (including
the unconsumed peeked line). Peeking line-per-line uses an internal
queue of lines that have been read ahead from the underlying stream.
The addition of the line-buffered reader is a workaround solution to
a problem with decoding. At current selecting a decoder works by
irreversibly reading the first line from the stream and looking for
a magic string that indicates the type of decoder to use.
It might however be possible for a file to be valid in format, just
missing a header. In such a case a lack of a line-buffered reader makes
it impossible to reparse the content of that first line. Introducing it
will however allow to peek the first line for magic first.
- If magic is found in the first line, GetDecoder() will peek it and
use it to return the correct Decoder instance. Note that in the case
of JsonBeatmapDecoder the magic is the opening JSON object brace,
and therefore must not be consumed.
- If magic is not found, the fallback decoder will be able to consume
it using ReadLine() in Decode().
This commit additionally contains basic unit tests for the reader.
Suggested-by: Aergwyn <aergwyn@t-online.de>
There is now a CreateContext method for retrieving a stand-alone context for threaded use.
We may want to add safety against this context being disposed (or just return a fresh one if it is).
Previously, in the rare case the database became out of sync with the disk store, it was impossible to feasibly repair a beatmap. Now reimporting checks each file exists on disk and adds it back if it doesn't.
