A corrupted file should be handled and raise a KeyError instead

count() should work the same as len(), but with the option to exclude expired responses

Test that if blake2b fails due to FIPS mode, it creates a valid key using a different hash function. Fallback on TypeError and ValueError - both a

All of the above variations should produce the same cache key

An error other than 'ttl already enabled' should be reraised

Run tests using the current virtualenv

This is just provided as an alias, since both requests and redirects share the same db file

Test that gzip-compressed responses read decompressed content with decode_content=True

Test that JSON responses (with both dict and list root) are correctly returned from the cache, regardless of `decode_content` setting

Test that JSON responses with an empty or primitive type root element are correctly returned from the cache, regardless of `decode_content` se

Test that other JSON media type variants are correctly decoded and returned from the cache

Common auth params and headers (for OAuth2, etc.) should be ignored by default

Some more tests to ensure ``delitem`` deletes only the expected items

Vacuum should not be used after delete if there is only a single key

Vacuum should not be used after delete if disabled

If a key does not exist, deleting it should not delete other keys.

DO_NOT_CACHE should bypass the cache on both read and write

If a file is too big, it should not be cached at all.

If a file is too big, it should not be cached at all.

Check that the files are added up to the size. After adding, the first ones are dropped.

An error other than 'table already exists' should be reraised

Test behahior when a file is being evicted but has already been (manually) deleted on disk

Check that the files are added up to the size. After adding, the least recently used files are removed.

CachedSession has an `expire_after` property for backwards-compatibility

EXPIRE_IMMEDIATELY should save a response only if it has a validator

Without stale_if_error (default), if there is an error while re-fetching an expired response, the request should be re-raised

Filtering by expired should use a more efficient SQL query

filter_fn should also apply to previously cached responses

An empty Expires header should raise ValueError, not IndexError

Compute the block size.

If there are multiple matches, the first match should be used in the order defined

Ensure that currently unimplemented Cache-Control headers do not affect behavior

include_get_headers is aliased to match_headers for backwards-compatibility

Test precedence with various combinations or per-request, per-session, per-URL, and Cache-Control expiration

This block size is invalid.

Cannot have file size greater than maximum cache size.

The argument is initialized properly.

Test passing the lock parameter.

Test with Cache-Control request headers

Test with Cache-Control request headers

Test with per-session, per-request, and per-URL expiration

Test behavior with both cache settings and request headers.

Test that the correct error is thrown when a user does not have a dependency installed

For a response that expired 10 seconds ago, it may be either accepted or rejected based on max-stale

For a response that expires in 10 seconds, it may be either accepted or rejected based on min-fresh

For a response that expired 10 seconds ago, if an error occurred while refreshing, it may be either accepted or rejected based on stale-if-error

For a response that expired 10 seconds ago, if an error occurred while refreshing, it may be either accepted or rejected based on stale-while-reva