1.0.11

Support manual skipping of subdirectories during indexing.

This provides an easy mechanism for excluding subdirectories from the search.
The intended use is to allow users to ignore older versions of a resource by
just creating a .SewerRatignore file in those versions' subdirectories.

1.0.10

Updated versions of all dependencies.

1.0.9

Use separate SQLite connection pools for reading and writing.

By only allowing a single write connection, we can leverage database/sql's
auto-blocking behavior to avoid SQLITE_BUSY errors on multiple write requests.

1.0.8

Centralize definition of timeout units in the argument parsing.

This avoids confusion over what the units are when Durations get passed
around. We also fix a bug in the exponential backoff for the polling.

1.0.7

Remove unnecessary argument from tokenizeMetadata's recursion.

1.0.6

Cleaned up the CLI option documentation.

1.0.5

Cleaned up the query text translator.

- Use only the first ':' for field discovery.
- Only switch to partial matches for the word that has '%'.
- Automatically create an AND clause for multi-word clauses.
- Added more tests, mention '.' field names in docs.

1.0.4

Skip hidden directories when listing metadata files for indexing. (#13)

This is primarily aimed at reducing the work required to process Git
repositories, though the same applies to other tools, e.g., .svn.

Added corresponding tests, along with some clean-ups to reduce
interdependencies between test scenarios.

Also streamlined docs to better explain the metadata listing process.

1.0.3

Process HEAD requests in the /retrieve/file handler.

This allows remote clients to check if the file has changed so that their
caches can be invalidated for a fresh download.

0.1.2

Cleaned up the README text.