yv01p/Umbraco-CMS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

U4-9121 - improve url perfs

Browse Source
This commit is contained in:
Stephan
2016-10-28 14:33:44 +02:00
parent 7495d89a79
commit c500f98ad8
6 changed files with 268 additions and 199 deletions
Download Patch File Download Diff File Expand all files Collapse all files

234
src/Umbraco.Web/PublishedCache/XmlPublishedCache/PublishedContentCache.cs
View File

@@ -16,7 +16,9 @@ using umbraco;
using System.Linq;
using umbraco.BusinessLogic;
using umbraco.presentation.preview;
using Umbraco.Core.Services;
using GlobalSettings = umbraco.GlobalSettings;
using Task = System.Threading.Tasks.Task;
namespace Umbraco.Web.PublishedCache.XmlPublishedCache
{
@@ -26,6 +28,13 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
private readonly RoutesCache _routesCache = new RoutesCache(!UnitTesting);
private DomainHelper _domainHelper;
private DomainHelper GetDomainHelper(IDomainService domainService)
{
return _domainHelper ?? (_domainHelper = new DomainHelper(domainService));
}
// for INTERNAL, UNIT TESTS use ONLY
internal RoutesCache RoutesCache { get { return _routesCache; } }
@@ -99,6 +108,13 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
// - non-colliding, adds one complete "by route" lookup, only on the first time a url is computed (then it's cached anyways)
// - colliding, adds one "by route" lookup, the first time the url is computed, then one dictionary looked each time it is computed again
// assuming no collisions, the impact is one complete "by route" lookup the first time each url is computed
//
// U4-9121 - this lookup is too expensive when computing a large amount of urls on a front-end (eg menu)
// ... thinking about moving the lookup out of the path into its own async task, so we are not reporting errors
// in the back-office anymore, but at least we are not polluting the cache
// instead, refactored DeterminedIdByRoute to stop using XPath, with a 16x improvement according to benchmarks
// will it be enough?
var loopId = preview ? 0 : _routesCache.GetNodeId(route); // might be cached already in case of collision
if (loopId == 0)
{
@@ -130,62 +146,141 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
var pos = route.IndexOf('/');
var path = pos == 0 ? route : route.Substring(pos);
var startNodeId = pos == 0 ? 0 : int.Parse(route.Substring(0, pos));
IEnumerable<XPathVariable> vars;
var xpath = CreateXpathQuery(startNodeId, path, hideTopLevelNode, out vars);
//check if we can find the node in our xml cache
var content = GetSingleByXPath(umbracoContext, preview, xpath, vars == null ? null : vars.ToArray());
var id = NavigateRoute(umbracoContext, preview, startNodeId, path, hideTopLevelNode);
if (id > 0) return GetById(umbracoContext, preview, id);
// if hideTopLevelNodePath is true then for url /foo we looked for /*/foo
// but maybe that was the url of a non-default top-level node, so we also
// have to look for /foo (see note in ApplyHideTopLevelNodeFromPath).
if (content == null && hideTopLevelNode && path.Length > 1 && path.IndexOf('/', 1) < 0)
if (hideTopLevelNode && path.Length > 1 && path.IndexOf('/', 1) < 0)
{
xpath = CreateXpathQuery(startNodeId, path, false, out vars);
content = GetSingleByXPath(umbracoContext, preview, xpath, vars == null ? null : vars.ToArray());
var id2 = NavigateRoute(umbracoContext, preview, startNodeId, path, false);
if (id2 > 0) return GetById(umbracoContext, preview, id2);
}
return content;
return null;
}
private int NavigateRoute(UmbracoContext umbracoContext, bool preview, int startNodeId, string path, bool hideTopLevelNode)
{
var xml = GetXml(umbracoContext, preview);
XmlElement elt;
// empty path
if (path == string.Empty || path == "/")
{
if (startNodeId > 0)
{
elt = xml.GetElementById(startNodeId.ToString(CultureInfo.InvariantCulture));
return elt == null ? -1 : startNodeId;
}
elt = null;
var min = int.MaxValue;
foreach (XmlElement e in xml.DocumentElement.ChildNodes)
{
var sortOrder = int.Parse(e.GetAttribute("sortOrder"));
if (sortOrder < min)
{
min = sortOrder;
elt = e;
}
}
return elt == null ? -1 : int.Parse(elt.GetAttribute("id"));
}
// non-empty path
elt = startNodeId <= 0
? xml.DocumentElement
: xml.GetElementById(startNodeId.ToString(CultureInfo.InvariantCulture));
if (elt == null) return -1;
var urlParts = path.Split(SlashChar, StringSplitOptions.RemoveEmptyEntries);
if (hideTopLevelNode && startNodeId <= 0)
{
foreach (XmlElement e in elt.ChildNodes)
{
var id = NavigateElementRoute(e, urlParts);
if (id > 0) return id;
}
return -1;
}
return NavigateElementRoute(elt, urlParts);
}
private static bool UseLegacySchema
{
get { return UmbracoConfig.For.UmbracoSettings().Content.UseLegacyXmlSchema; }
}
private int NavigateElementRoute(XmlElement elt, string[] urlParts)
{
var found = true;
var i = 0;
while (found && i < urlParts.Length)
{
found = false;
foreach (XmlElement child in elt.ChildNodes)
{
var noNode = UseLegacySchema
? child.Name != "node"
: child.GetAttributeNode("isDoc") == null;
if (noNode) continue;
if (child.GetAttribute("urlName") != urlParts[i]) continue;
found = true;
elt = child;
break;
}
i++;
}
return found ? int.Parse(elt.GetAttribute("id")) : -1;
}
string DetermineRouteById(UmbracoContext umbracoContext, bool preview, int contentId)
{
var node = GetById(umbracoContext, preview, contentId);
if (node == null)
return null;
var elt = GetXml(umbracoContext, preview).GetElementById(contentId.ToString(CultureInfo.InvariantCulture));
if (elt == null) return null;
var domainHelper = new DomainHelper(umbracoContext.Application.Services.DomainService);
var domainHelper = GetDomainHelper(umbracoContext.Application.Services.DomainService);
// walk up from that node until we hit a node with a domain,
// or we reach the content root, collecting urls in the way
var pathParts = new List<string>();
var n = node;
var hasDomains = domainHelper.NodeHasDomains(n.Id);
while (hasDomains == false && n != null) // n is null at root
var eltId = int.Parse(elt.GetAttribute("id"));
var eltParentId = int.Parse(((XmlElement) elt.ParentNode).GetAttribute("id"));
var e = elt;
var id = eltId;
var hasDomains = domainHelper.NodeHasDomains(id);
while (hasDomains == false && id != -1)
{
// get the url
var urlName = n.UrlName;
var urlName = e.GetAttribute("urlName");
pathParts.Add(urlName);
// move to parent node
n = n.Parent;
hasDomains = n != null && domainHelper.NodeHasDomains(n.Id);
e = (XmlElement) e.ParentNode;
id = int.Parse(e.GetAttribute("id"));
hasDomains = id != -1 && domainHelper.NodeHasDomains(id);
}
// no domain, respect HideTopLevelNodeFromPath for legacy purposes
if (hasDomains == false && global::umbraco.GlobalSettings.HideTopLevelNodeFromPath)
ApplyHideTopLevelNodeFromPath(umbracoContext, node, pathParts);
if (hasDomains == false && GlobalSettings.HideTopLevelNodeFromPath)
ApplyHideTopLevelNodeFromPath(umbracoContext, eltId, eltParentId, pathParts);
// assemble the route
pathParts.Reverse();
var path = "/" + string.Join("/", pathParts); // will be "/" or "/foo" or "/foo/bar" etc
var route = (n == null ? "" : n.Id.ToString(CultureInfo.InvariantCulture)) + path;
var route = (id == -1 ? "" : id.ToString(CultureInfo.InvariantCulture)) + path;
return route;
}
static void ApplyHideTopLevelNodeFromPath(UmbracoContext umbracoContext, IPublishedContent node, IList<string> pathParts)
static void ApplyHideTopLevelNodeFromPath(UmbracoContext umbracoContext, int nodeId, int parentId, IList<string> pathParts)
{
// in theory if hideTopLevelNodeFromPath is true, then there should be only once
// top-level node, or else domains should be assigned. but for backward compatibility
@@ -195,12 +290,12 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
// "/foo" fails (looking for "/*/foo") we try also "/foo".
// this does not make much sense anyway esp. if both "/foo/" and "/bar/foo" exist, but
// that's the way it works pre-4.10 and we try to be backward compat for the time being
if (node.Parent == null)
if (parentId == -1)
{
var rootNode = umbracoContext.ContentCache.GetByRoute("/", true);
if (rootNode == null)
throw new Exception("Failed to get node at /.");
if (rootNode.Id == node.Id) // remove only if we're the default node
if (rootNode.Id == nodeId) // remove only if we're the default node
pathParts.RemoveAt(pathParts.Count - 1);
}
else
@@ -217,12 +312,7 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
{
public int Version { get; private set; }
public static string Root { get { return "/root"; } }
public string RootDocuments { get; private set; }
public string DescendantDocumentById { get; private set; }
public string ChildDocumentByUrlName { get; private set; }
public string ChildDocumentByUrlNameVar { get; private set; }
public string RootDocumentWithLowestSortOrder { get; private set; }
public XPathStringsDefinition(int version)
{
@@ -233,19 +323,11 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
// legacy XML schema
case 0:
RootDocuments = "/root/node";
DescendantDocumentById = "//node [@id={0}]";
ChildDocumentByUrlName = "/node [@urlName='{0}']";
ChildDocumentByUrlNameVar = "/node [@urlName=${0}]";
RootDocumentWithLowestSortOrder = "/root/node [not(@sortOrder > ../node/@sortOrder)][1]";
break;
// default XML schema as of 4.10
case 1:
RootDocuments = "/root/* [@isDoc]";
DescendantDocumentById = "//* [@isDoc and @id={0}]";
ChildDocumentByUrlName = "/* [@isDoc and @urlName='{0}']";
ChildDocumentByUrlNameVar = "/* [@isDoc and @urlName=${0}]";
RootDocumentWithLowestSortOrder = "/root/* [@isDoc and not(@sortOrder > ../* [@isDoc]/@sortOrder)][1]";
break;
default:
@@ -421,84 +503,6 @@ namespace Umbraco.Web.PublishedCache.XmlPublishedCache
static readonly char[] SlashChar = new[] { '/' };
protected string CreateXpathQuery(int startNodeId, string path, bool hideTopLevelNodeFromPath, out IEnumerable<XPathVariable> vars)
{
string xpath;
vars = null;
if (path == string.Empty || path == "/")
{
// if url is empty
if (startNodeId > 0)
{
// if in a domain then use the root node of the domain
xpath = string.Format(XPathStringsDefinition.Root + XPathStrings.DescendantDocumentById, startNodeId);
}
else
{
// if not in a domain - what is the default page?
// let's say it is the first one in the tree, if any -- order by sortOrder
// but!
// umbraco does not consistently guarantee that sortOrder starts with 0
// so the one that we want is the one with the smallest sortOrder
// read http://stackoverflow.com/questions/1128745/how-can-i-use-xpath-to-find-the-minimum-value-of-an-attribute-in-a-set-of-elemen
// so that one does not work, because min(@sortOrder) maybe 1
// xpath = "/root/*[@isDoc and @sortOrder='0']";
// and we can't use min() because that's XPath 2.0
// that one works
xpath = XPathStrings.RootDocumentWithLowestSortOrder;
}
}
else
{
// if url is not empty, then use it to try lookup a matching page
var urlParts = path.Split(SlashChar, StringSplitOptions.RemoveEmptyEntries);
var xpathBuilder = new StringBuilder();
int partsIndex = 0;
List<XPathVariable> varsList = null;
if (startNodeId == 0)
{
if (hideTopLevelNodeFromPath)
xpathBuilder.Append(XPathStrings.RootDocuments); // first node is not in the url
else
xpathBuilder.Append(XPathStringsDefinition.Root);
}
else
{
xpathBuilder.AppendFormat(XPathStringsDefinition.Root + XPathStrings.DescendantDocumentById, startNodeId);
// always "hide top level" when there's a domain
}
while (partsIndex < urlParts.Length)
{
var part = urlParts[partsIndex++];
if (part.Contains('\'') || part.Contains('"'))
{
// use vars, escaping gets ugly pretty quickly
varsList = varsList ?? new List<XPathVariable>();
var varName = string.Format("var{0}", partsIndex);
varsList.Add(new XPathVariable(varName, part));
xpathBuilder.AppendFormat(XPathStrings.ChildDocumentByUrlNameVar, varName);
}
else
{
xpathBuilder.AppendFormat(XPathStrings.ChildDocumentByUrlName, part);
}
}
xpath = xpathBuilder.ToString();
if (varsList != null)
vars = varsList.ToArray();
}
return xpath;
}
#endregion
#region Detached