dircache: simplify interface, fix corner cases and apply to backends

Dircache was changed to:

- Remove special cases for the root directory
- Remove Fatal errors
- Call FindRoot on behalf of the user wherever possible
- Bring up to modern Go standards

Backends were changed to:

- Remove calls to FindRoot
- Change calls to FindRootAndPath to FindPath
- Don't make special cases for the root

This fixes several corner cases, for example removing a non existent
directory if FindRoot hasn't been called.
This commit is contained in:
Nick Craig-Wood
2020-05-11 17:24:37 +01:00
parent 81151523af
commit a8652e2252
14 changed files with 234 additions and 633 deletions

View File

@@ -1,4 +1,5 @@
// Package dircache provides a simple cache for caching directory to path lookups
// Package dircache provides a simple cache for caching directory ID
// to path lookups and the inverse.
package dircache
// _methods are called without the lock
@@ -7,7 +8,6 @@ import (
"bytes"
"context"
"fmt"
"log"
"strings"
"sync"
@@ -17,19 +17,23 @@ import (
// DirCache caches paths to directory IDs and vice versa
type DirCache struct {
cacheMu sync.RWMutex
cache map[string]string
invCache map[string]string
mu sync.Mutex
fs DirCacher // Interface to find and make stuff
trueRootID string // ID of the absolute root
root string // the path we are working on
rootID string // ID of the root directory
rootParentID string // ID of the root's parent directory
foundRoot bool // Whether we have found the root or not
cacheMu sync.RWMutex // protects cache and invCache
cache map[string]string
invCache map[string]string
mu sync.Mutex // protects the below
fs DirCacher // Interface to find and make directories
trueRootID string // ID of the absolute root
root string // the path the cache is rooted on
rootID string // ID of the root directory
rootParentID string // ID of the root's parent directory
foundRoot bool // Whether we have found the root or not
}
// DirCacher describes an interface for doing the low level directory work
//
// This should be implemented by the backend and will be called by the
// dircache package when appropriate.
type DirCacher interface {
FindLeaf(ctx context.Context, pathID, leaf string) (pathIDOut string, found bool, err error)
CreateDir(ctx context.Context, pathID, leaf string) (newID string, err error)
@@ -37,6 +41,15 @@ type DirCacher interface {
// New makes a DirCache
//
// This is created with the true root ID and the root path.
//
// In order to use the cache FindRoot() must be called on it without
// error. This isn't done at initialization as it isn't known whether
// the root and intermediate directories need to be created or not.
//
// Most of the utility functions wil call FindRoot() on the caller's
// behalf with the create flag passed in.
//
// The cache is safe for concurrent use
func New(root string, trueRootID string, fs DirCacher) *DirCache {
d := &DirCache{
@@ -74,23 +87,29 @@ func (dc *DirCache) String() string {
return buf.String()
}
// Get an ID given a path
// Get a directory ID given a path
//
// Returns the ID and a boolean as to whether it was found or not in
// the cache.
func (dc *DirCache) Get(path string) (id string, ok bool) {
dc.cacheMu.RLock()
id, ok = dc.cache[path]
dc.cacheMu.RUnlock()
return
return id, ok
}
// GetInv gets a path given an ID
// GetInv gets a path given a directory ID
//
// Returns the path and a boolean as to whether it was found or not in
// the cache.
func (dc *DirCache) GetInv(id string) (path string, ok bool) {
dc.cacheMu.RLock()
path, ok = dc.invCache[id]
dc.cacheMu.RUnlock()
return
return path, ok
}
// Put a path, id into the map
// Put a (path, directory ID) pair into the cache
func (dc *DirCache) Put(path, id string) {
dc.cacheMu.Lock()
dc.cache[path] = id
@@ -98,7 +117,7 @@ func (dc *DirCache) Put(path, id string) {
dc.cacheMu.Unlock()
}
// Flush the map of all data
// Flush the cache of all data
func (dc *DirCache) Flush() {
dc.cacheMu.Lock()
dc.cache = make(map[string]string)
@@ -106,9 +125,10 @@ func (dc *DirCache) Flush() {
dc.cacheMu.Unlock()
}
// FlushDir flushes the map of all data starting with dir
// FlushDir flushes the map of all data starting with with the path
// dir.
//
// If dir is empty then this is equivalent to calling ResetRoot
// If dir is empty string then this is equivalent to calling ResetRoot
func (dc *DirCache) FlushDir(dir string) {
if dir == "" {
dc.ResetRoot()
@@ -159,39 +179,29 @@ func SplitPath(path string) (directory, leaf string) {
//
// If create is set it will make the directory if not found
//
// Algorithm:
// Look in the cache for the path, if found return the pathID
// If not found strip the last path off the path and recurse
// Now have a parent directory id, so look in the parent for self and return it
// It will call FindRoot if it hasn't been called already
func (dc *DirCache) FindDir(ctx context.Context, path string, create bool) (pathID string, err error) {
dc.mu.Lock()
defer dc.mu.Unlock()
err = dc._findRoot(ctx, create)
if err != nil {
return "", err
}
return dc._findDir(ctx, path, create)
}
// Look for the root and in the cache - safe to call without the mu
func (dc *DirCache) _findDirInCache(path string) string {
// fmt.Println("Finding",path,"create",create,"cache",cache)
// Unlocked findDir
//
// Call with a lock on mu
func (dc *DirCache) _findDir(ctx context.Context, path string, create bool) (pathID string, err error) {
// If it is the root, then return it
if path == "" {
// fmt.Println("Root")
return dc.rootID
return dc.rootID, nil
}
// If it is in the cache then return it
pathID, ok := dc.Get(path)
if ok {
// fmt.Println("Cache hit on", path)
return pathID
}
return ""
}
// Unlocked findDir - must have mu
func (dc *DirCache) _findDir(ctx context.Context, path string, create bool) (pathID string, err error) {
pathID = dc._findDirInCache(path)
if pathID != "" {
return pathID, nil
}
@@ -232,29 +242,48 @@ func (dc *DirCache) _findDir(ctx context.Context, path string, create bool) (pat
// FindPath finds the leaf and directoryID from a path
//
// Do not call FindPath with the root directory - it will return an error
// If called with path == "" then it will return the ID of the parent
// directory of the root and the leaf name of the root in that
// directory. Note that it won't create the root directory in this
// case even if create is true.
//
// If create is set parent directories will be created if they don't exist
//
// It will call FindRoot if it hasn't been called already
func (dc *DirCache) FindPath(ctx context.Context, path string, create bool) (leaf, directoryID string, err error) {
if path == "" {
err = errors.New("internal error: can't call FindPath with root directory")
return
_, leaf = SplitPath(dc.root)
directoryID, err = dc.RootParentID(ctx, create)
} else {
var directory string
directory, leaf = SplitPath(path)
directoryID, err = dc.FindDir(ctx, directory, create)
}
dc.mu.Lock()
defer dc.mu.Unlock()
directory, leaf := SplitPath(path)
directoryID, err = dc._findDir(ctx, directory, create)
return
return leaf, directoryID, err
}
// FindRoot finds the root directory if not already found
//
// If successful this changes the root of the cache from the true root
// to the root specified by the path passed into New.
//
// Resets the root directory
//
// If create is set it will make the directory if not found
func (dc *DirCache) FindRoot(ctx context.Context, create bool) error {
dc.mu.Lock()
defer dc.mu.Unlock()
return dc._findRoot(ctx, create)
}
// _findRoot finds the root directory if not already found
//
// Resets the root directory
//
// If create is set it will make the directory if not found
//
// Call with mu held
func (dc *DirCache) _findRoot(ctx context.Context, create bool) error {
if dc.foundRoot {
return nil
}
@@ -277,51 +306,51 @@ func (dc *DirCache) FindRoot(ctx context.Context, create bool) error {
return nil
}
// FindRootAndPath finds the root first if not found then finds leaf and directoryID from a path
//
// If create is set parent directories will be created if they don't exist
func (dc *DirCache) FindRootAndPath(ctx context.Context, path string, create bool) (leaf, directoryID string, err error) {
err = dc.FindRoot(ctx, create)
if err != nil {
return
}
return dc.FindPath(ctx, path, create)
}
// FoundRoot returns whether the root directory has been found yet
//
// Call this from FindLeaf or CreateDir only
func (dc *DirCache) FoundRoot() bool {
dc.mu.Lock()
defer dc.mu.Unlock()
return dc.foundRoot
}
// RootID returns the ID of the root directory
//
// This should be called after FindRoot
func (dc *DirCache) RootID() string {
// If create is set it will make the root directory if not found
func (dc *DirCache) RootID(ctx context.Context, create bool) (ID string, err error) {
dc.mu.Lock()
defer dc.mu.Unlock()
if !dc.foundRoot {
log.Fatalf("Internal Error: RootID() called before FindRoot")
err = dc._findRoot(ctx, create)
if err != nil {
return "", err
}
return dc.rootID
return dc.rootID, nil
}
// RootParentID returns the ID of the parent of the root directory
//
// This should be called after FindRoot
func (dc *DirCache) RootParentID() (string, error) {
// If create is set it will make the root parent directory if not found (but not the root)
func (dc *DirCache) RootParentID(ctx context.Context, create bool) (ID string, err error) {
dc.mu.Lock()
defer dc.mu.Unlock()
if !dc.foundRoot {
return "", errors.New("internal error: RootID() called before FindRoot")
if dc.root == "" {
return "", errors.New("is root directory")
}
// Find the rootParentID without creating the root
rootParent, _ := SplitPath(dc.root)
rootParentID, err := dc._findDir(ctx, rootParent, create)
if err != nil {
return "", err
}
dc.rootParentID = rootParentID
} else {
if dc.rootID == dc.trueRootID {
return "", errors.New("is root directory")
}
}
if dc.rootParentID == "" {
return "", errors.New("internal error: didn't find rootParentID")
}
if dc.rootID == dc.trueRootID {
return "", errors.New("is root directory")
}
return dc.rootParentID, nil
}