Restructure data layer (#2532)

* Add new txn manager interface
* Add txn management to sqlite
* Rename get to getByID
* Add contexts to repository methods
* Update query builders
* Add context to reader writer interfaces
* Use repository in resolver
* Tighten interfaces
* Tighten interfaces in dlna
* Tighten interfaces in match package
* Tighten interfaces in scraper package
* Tighten interfaces in scan code
* Tighten interfaces on autotag package
* Remove ReaderWriter usage
* Merge database package into sqlite

pkg/sqlite/regex.go

@@ -0,0 +1,42 @@
+package sqlite
+
+import (
+	"regexp"
+
+	lru "github.com/hashicorp/golang-lru"
+)
+
+// size of the regex LRU cache in elements.
+// A small number number was chosen because it's most likely use is for a
+// single query - this function gets called for every row in the (filtered)
+// results. It's likely to only need no more than 1 or 2 in any given query.
+// After that point, it's just sitting in the cache and is unlikely to be used
+// again.
+const regexCacheSize = 10
+
+var regexCache *lru.Cache
+
+func init() {
+	regexCache, _ = lru.New(regexCacheSize)
+}
+
+// regexFn is registered as an SQLite function as "regexp"
+// It uses an LRU cache to cache recent regex patterns to reduce CPU load over
+// identical patterns.
+func regexFn(re, s string) (bool, error) {
+	entry, ok := regexCache.Get(re)
+	var compiled *regexp.Regexp
+
+	if !ok {
+		var err error
+		compiled, err = regexp.Compile(re)
+		if err != nil {
+			return false, err
+		}
+		regexCache.Add(re, compiled)
+	} else {
+		compiled = entry.(*regexp.Regexp)
+	}
+
+	return compiled.MatchString(s), nil
+}