public async Task <List <ChargePoint> > DeDuplicateList(List <ChargePoint> cpList, bool updateDuplicate, CoreReferenceData coreRefData, ImportReport report, bool allowDupeWithDifferentOperator = false, bool fetchExistingFromAPI = false)
{
var stopWatch = new Stopwatch();
stopWatch.Start();
var poiManager = new POIManager();
//get list of all current POIs (in relevant countries) including most delisted ones
int[] countryIds = (from poi in cpList
where poi.AddressInfo.Country != null
select poi.AddressInfo.Country.ID).Distinct().ToArray();
APIRequestParams filters = new APIRequestParams {
CountryIDs = countryIds, MaxResults = 1000000, EnableCaching = true, SubmissionStatusTypeID = 0
};
List <ChargePoint> masterListCollection = new List <ChargePoint>();
if (fetchExistingFromAPI)
{
// fetch from API
masterListCollection = await new OCMClient(false).GetLocations(new SearchFilters
{
CountryIDs = countryIds,
MaxResults = 1000000,
EnableCaching = true,
SubmissionStatusTypeIDs = new int[0]
});
}
else
{
// use local database
masterListCollection = poiManager.GetChargePoints(filters);
}
var spec = new i4o.IndexSpecification <ChargePoint>()
.Add(i => i.DataProviderID)
.Add(i => i.DataProvidersReference)
;
var masterList = new i4o.IndexSet <ChargePoint>(masterListCollection, spec);
List <ChargePoint> masterListCopy = new List <ChargePoint>();
foreach (var tmp in masterList)
{
//fully copy of master list item so we have before/after
masterListCopy.Add(JsonConvert.DeserializeObject <ChargePoint>(JsonConvert.SerializeObject(tmp)));
}
//if we failed to get a master list, quit with no result
if (masterListCollection.Count == 0)
{
return(new List <ChargePoint>());
}
List <ChargePoint> duplicateList = new List <ChargePoint>();
List <ChargePoint> updateList = new List <ChargePoint>();
ChargePoint previousCP = null;
//for each item to be imported, deduplicate by adding to updateList only the items which we don't already haves
var cpListSortedByPos = cpList.OrderBy(c => c.AddressInfo.Latitude).ThenBy(c => c.AddressInfo.Longitude);
int poiProcessed = 0;
int totalPOI = cpListSortedByPos.Count();
Stopwatch dupeIdentWatch = new Stopwatch();
dupeIdentWatch.Start();
foreach (var item in cpListSortedByPos)
{
var itemGeoPos = new System.Device.Location.GeoCoordinate(item.AddressInfo.Latitude, item.AddressInfo.Longitude);
//item is duplicate if we already seem to have it based on Data Providers reference or approx position match
var dupeList = masterList.Where(c =>
(
// c.DataProvider != null &&
c.DataProviderID == item.DataProviderID && c.DataProvidersReference == item.DataProvidersReference) ||
(c.AddressInfo.Title == item.AddressInfo.Title && c.AddressInfo.AddressLine1 == item.AddressInfo.AddressLine1 && c.AddressInfo.Postcode == item.AddressInfo.Postcode) ||
(GeoManager.IsClose(c.AddressInfo.Latitude, c.AddressInfo.Longitude, item.AddressInfo.Latitude, item.AddressInfo.Longitude) && new System.Device.Location.GeoCoordinate(c.AddressInfo.Latitude, c.AddressInfo.Longitude).GetDistanceTo(itemGeoPos) < DUPLICATE_DISTANCE_METERS) //meters distance apart
);
if (dupeList.Any())
{
if (updateDuplicate)
{
//if updating duplicates, get exact matching duplicate based on provider reference and update/merge with this item to update status/merge properties
var updatedItem = dupeList.FirstOrDefault(d => d.DataProviderID == (item.DataProvider != null ? item.DataProvider.ID : item.DataProviderID) && d.DataProvidersReference == item.DataProvidersReference);
if (updatedItem != null)
{
//only merge/update from live published items
if (updatedItem.SubmissionStatus.IsLive == (bool?)true ||
updatedItem.SubmissionStatus.ID == (int)StandardSubmissionStatusTypes.Delisted_RemovedByDataProvider ||
updatedItem.SubmissionStatus.ID == (int)StandardSubmissionStatusTypes.Delisted_NotPublicInformation)
{
//item is an exact match from same data provider
//overwrite existing with imported data (use import as master)
//updatedItem = poiManager.PreviewPopulatedPOIFromModel(updatedItem);
MergeItemChanges(item, updatedItem, false);
updateList.Add(updatedItem);
}
}
if (updatedItem == null)
{
//duplicates are not exact match
//TODO: resolve whether imported data should change duplicate
//merge new properties from imported item
//if (item.StatusType != null) updatedItem.StatusType = item.StatusType;
//updateList.Add(updatedItem);
}
}
//item has one or more likely duplicates, add it to list of items to remove
duplicateList.Add(item);
}
//mark item as duplicate if location/title exactly matches previous entry or lat/long is within DuplicateDistance meters
if (previousCP != null)
{
//this branch is the most expensive part of dedupe:
if (IsDuplicateLocation(item, previousCP, true))
{
if (!duplicateList.Contains(item))
{
if (allowDupeWithDifferentOperator && item.OperatorID != previousCP.OperatorID)
{
Log("Duplicated allowed due to different operator:" + item.AddressInfo.Title);
}
else
{
Log("Duplicated item removed:" + item.AddressInfo.Title);
duplicateList.Add(item);
}
}
}
}
previousCP = item;
poiProcessed++;
if (poiProcessed % 300 == 0)
{
System.Diagnostics.Debug.WriteLine("Deduplication: " + poiProcessed + " processed of " + totalPOI);
}
}
dupeIdentWatch.Stop();
Log("De-dupe pass took " + dupeIdentWatch.Elapsed.TotalSeconds + " seconds. " + (dupeIdentWatch.Elapsed.TotalMilliseconds / cpList.Count) + "ms per item.");
//remove duplicates from list to apply
foreach (var dupe in duplicateList)
{
cpList.Remove(dupe);
}
Log("Duplicates removed from import:" + duplicateList.Count);
//add updated items (replace duplicates with property changes)
foreach (var updatedItem in updateList)
{
if (!cpList.Contains(updatedItem))
{
cpList.Add(updatedItem);
}
}
Log("Updated items to import:" + updateList.Count);
//populate missing location info from geolocation cache if possible
Stopwatch geoWatch = new Stopwatch();
geoWatch.Start();
PopulateLocationFromGeolocationCache(cpList, coreRefData);
geoWatch.Stop();
Log("Populate Country from Lat/Long took " + geoWatch.Elapsed.TotalSeconds + " seconds. " + (geoWatch.Elapsed.TotalMilliseconds / cpList.Count) + "ms per item.");
//final pass to catch duplicates present in data source, mark additional items as Delisted Duplicate so we have a record for them
var submissionStatusDelistedDupe = coreRefData.SubmissionStatusTypes.First(s => s.ID == 1001); //delisted duplicate
previousCP = null;
//sort current cp list by position again
cpListSortedByPos = cpList.OrderBy(c => c.AddressInfo.Latitude).ThenBy(c => c.AddressInfo.Longitude);
//mark any duplicates in final list as delisted duplicates (submitted to api)
foreach (var cp in cpListSortedByPos)
{
bool isDuplicate = false;
if (previousCP != null)
{
isDuplicate = IsDuplicateLocation(cp, previousCP, false);
if (isDuplicate)
{
cp.SubmissionStatus = submissionStatusDelistedDupe;
cp.SubmissionStatusTypeID = submissionStatusDelistedDupe.ID;
if (previousCP.ID > 0)
{
if (cp.GeneralComments == null)
{
cp.GeneralComments = "";
}
cp.GeneralComments += " [Duplicate of OCM-" + previousCP.ID + "]";
cp.ParentChargePointID = previousCP.ID;
}
}
}
if (!isDuplicate)
{
previousCP = cp;
}
}
report.Added = cpListSortedByPos.Where(cp => cp.ID == 0).ToList();
report.Updated = cpListSortedByPos.Where(cp => cp.ID > 0).ToList();
report.Duplicates = duplicateList; //TODO: add additional pass of duplicates from above
//determine which POIs in our master list are no longer referenced in the import
report.Delisted = masterList.Where(cp => cp.DataProviderID == report.ProviderDetails.DataProviderID && cp.SubmissionStatus != null && (cp.SubmissionStatus.IsLive == true || cp.SubmissionStatusTypeID == (int)StandardSubmissionStatusTypes.Imported_UnderReview) &&
!cpListSortedByPos.Any(master => master.ID == cp.ID) && !report.Duplicates.Any(master => master.ID == cp.ID) &&
cp.UserComments == null && cp.MediaItems == null).ToList();
//safety check to ensure we're not delisting items just because we have incomplete import data:
if (cpList.Count < 50)// || (report.Delisted.Count > cpList.Count))
{
report.Delisted = new List <ChargePoint>();
}
//determine list of low quality POIs (incomplete address info etc)
report.LowDataQuality = new List <ChargePoint>();
report.LowDataQuality.AddRange(GetLowDataQualityPOIs(report.Added));
report.LowDataQuality.AddRange(GetLowDataQualityPOIs(report.Updated));
Log("Removing " + report.LowDataQuality.Count + " low quality POIs from added/updated");
//remove references in added/updated to any low quality POIs
foreach (var p in report.LowDataQuality)
{
report.Added.Remove(p);
}
foreach (var p in report.LowDataQuality)
{
report.Updated.Remove(p);
}
//remove updates which only change datelaststatusupdate
var updatesToIgnore = new List <ChargePoint>();
foreach (var poi in report.Updated)
{
var origPOI = masterListCopy.FirstOrDefault(p => p.ID == poi.ID);
var updatedPOI = poiManager.PreviewPopulatedPOIFromModel(poi);
var differences = poiManager.CheckDifferences(origPOI, updatedPOI);
differences.RemoveAll(d => d.Context == ".MetadataValues");
differences.RemoveAll(d => d.Context == ".DateLastStatusUpdate");
differences.RemoveAll(d => d.Context == ".UUID");
differences.RemoveAll(d => d.Context == ".DataProvider.DateLastImported");
differences.RemoveAll(d => d.Context == ".IsRecentlyVerified");
differences.RemoveAll(d => d.Context == ".DateLastVerified");
differences.RemoveAll(d => d.Context == ".UserComments");
differences.RemoveAll(d => d.Context == ".MediaItems");
if (!differences.Any())
{
updatesToIgnore.Add(poi);
}
else
{
//differences exist
CompareLogic compareLogic = new CompareLogic();
compareLogic.Config.MaxDifferences = 100;
compareLogic.Config.IgnoreObjectTypes = false;
compareLogic.Config.IgnoreUnknownObjectTypes = true;
compareLogic.Config.CompareChildren = true;
ComparisonResult result = compareLogic.Compare(origPOI, updatedPOI);
var diffReport = new KellermanSoftware.CompareNetObjects.Reports.UserFriendlyReport();
result.Differences.RemoveAll(d => d.PropertyName == ".MetadataValues");
result.Differences.RemoveAll(d => d.PropertyName == ".DateLastStatusUpdate");
result.Differences.RemoveAll(d => d.PropertyName == ".UUID");
result.Differences.RemoveAll(d => d.PropertyName == ".DataProvider.DateLastImported");
result.Differences.RemoveAll(d => d.PropertyName == ".IsRecentlyVerified");
result.Differences.RemoveAll(d => d.PropertyName == ".DateLastVerified");
result.Differences.RemoveAll(d => d.PropertyName == ".UserComments");
result.Differences.RemoveAll(d => d.PropertyName == ".MediaItems");
System.Diagnostics.Debug.WriteLine("Difference:" + diffReport.OutputString(result.Differences));
if (!result.Differences.Any())
{
updatesToIgnore.Add(poi);
}
}
}
foreach (var p in updatesToIgnore)
{
if (report.Unchanged == null)
{
report.Unchanged = new List <ChargePoint>();
}
report.Unchanged.Add(p);
report.Updated.Remove(p);
}
//TODO: if POi is a duplicate ensure imported data provider reference/URL is included as reference metadata in OCM's version of the POI
stopWatch.Stop();
Log("Deduplicate List took " + stopWatch.Elapsed.TotalSeconds + " seconds");
//return final processed list ready for applying as insert/updates
return(cpListSortedByPos.ToList());
}
/**
* Generic Import Process
Provider Properties
Import Method
Import URL/Path
Import Frequency
IsMaster
Fetch Latest Data
For each item
Check If Exists or Strong Duplicate, Get ID
If New, Add
if Exists Then
Prepare update, if provider supports live status, set that
What if item updated manually on OCM?
Send Update
End
Loop
Log Exceptions
Log Count of Items Added or Modified
Way to remove item (or log items) which no longer exist in data source?
* */
public async Task<List<ChargePoint>> DeDuplicateList(List<ChargePoint> cpList, bool updateDuplicate, CoreReferenceData coreRefData, ImportReport report, bool allowDupeWithDifferentOperator = false)
{
var stopWatch = new Stopwatch();
stopWatch.Start();
//get list of all current POIs (in relevant countries) including most delisted ones
int[] countryIds = (from poi in cpList
where poi.AddressInfo.Country != null
select poi.AddressInfo.Country.ID).Distinct().ToArray();
APIRequestParams filters = new APIRequestParams { CountryIDs = countryIds, MaxResults = 1000000, EnableCaching = false, SubmissionStatusTypeID = 0 };
//List<ChargePoint> masterList = await new OCMClient(IsSandboxedAPIMode).GetLocations(filters); //new OCMClient().FindSimilar(null, 10000); //fetch all charge points regardless of status
var poiManager = new POIManager();
List<ChargePoint> masterListCollection = poiManager.GetChargePoints(filters); //new OCMClient().FindSimilar(null, 10000); //fetch all charge points regardless of status
var spec = new i4o.IndexSpecification<ChargePoint>()
.Add(i => i.DataProviderID)
.Add(i => i.DataProvidersReference)
;
var masterList = new i4o.IndexSet<ChargePoint>(masterListCollection, spec);
List<ChargePoint> masterListCopy = new List<ChargePoint>();
foreach (var tmp in masterList)
{
//fully copy of master list item so we have before/after
masterListCopy.Add(JsonConvert.DeserializeObject<ChargePoint>(JsonConvert.SerializeObject(tmp)));
}
//if we failed to get a master list, quit with no result
if (masterListCollection.Count == 0) return new List<ChargePoint>();
List<ChargePoint> duplicateList = new List<ChargePoint>();
List<ChargePoint> updateList = new List<ChargePoint>();
ChargePoint previousCP = null;
//for each item to be imported, deduplicate by adding to updateList only the items which we don't already haves
var cpListSortedByPos = cpList.OrderBy(c => c.AddressInfo.Latitude).ThenBy(c => c.AddressInfo.Longitude);
int poiProcessed = 0;
int totalPOI = cpListSortedByPos.Count();
Stopwatch dupeIdentWatch = new Stopwatch();
dupeIdentWatch.Start();
foreach (var item in cpListSortedByPos)
{
var itemGeoPos = new System.Device.Location.GeoCoordinate(item.AddressInfo.Latitude, item.AddressInfo.Longitude);
//item is duplicate if we already seem to have it based on Data Providers reference or approx position match
var dupeList = masterList.Where(c =>
(c.DataProvider != null && c.DataProviderID == item.DataProviderID && c.DataProvidersReference == item.DataProvidersReference)
|| (c.AddressInfo.Title == item.AddressInfo.Title && c.AddressInfo.AddressLine1 == item.AddressInfo.AddressLine1 && c.AddressInfo.Postcode == item.AddressInfo.Postcode)
|| (GeoManager.IsClose(c.AddressInfo.Latitude, c.AddressInfo.Longitude, item.AddressInfo.Latitude, item.AddressInfo.Longitude) && new System.Device.Location.GeoCoordinate(c.AddressInfo.Latitude, c.AddressInfo.Longitude).GetDistanceTo(itemGeoPos) < DUPLICATE_DISTANCE_METERS) //meters distance apart
);
if (dupeList.Any())
{
if (updateDuplicate)
{
//if updating duplicates, get exact matching duplicate based on provider reference and update/merge with this item to update status/merge properties
var updatedItem = dupeList.FirstOrDefault(d => d.DataProviderID == (item.DataProvider != null ? item.DataProvider.ID : item.DataProviderID) && d.DataProvidersReference == item.DataProvidersReference);
if (updatedItem != null)
{
//only merge/update from live published items
if (updatedItem.SubmissionStatus.IsLive == (bool?)true
|| updatedItem.SubmissionStatus.ID == (int)StandardSubmissionStatusTypes.Delisted_RemovedByDataProvider
|| updatedItem.SubmissionStatus.ID == (int)StandardSubmissionStatusTypes.Delisted_NotPublicInformation)
{
//item is an exact match from same data provider
//overwrite existing with imported data (use import as master)
//updatedItem = poiManager.PreviewPopulatedPOIFromModel(updatedItem);
MergeItemChanges(item, updatedItem, false);
updateList.Add(updatedItem);
}
}
if (updatedItem == null)
{
//duplicates are not exact match
//TODO: resolve whether imported data should change duplicate
//merge new properties from imported item
//if (item.StatusType != null) updatedItem.StatusType = item.StatusType;
//updateList.Add(updatedItem);
}
}
//item has one or more likely duplicates, add it to list of items to remove
duplicateList.Add(item);
}
//mark item as duplicate if location/title exactly matches previous entry or lat/long is within DuplicateDistance meters
if (previousCP != null)
{
//this branch is the most expensive part of dedupe:
if (IsDuplicateLocation(item, previousCP, true))
{
if (!duplicateList.Contains(item))
{
if (allowDupeWithDifferentOperator && item.OperatorID != previousCP.OperatorID)
{
Log("Duplicated allowed due to different operator:" + item.AddressInfo.Title);
}
else
{
Log("Duplicated item removed:" + item.AddressInfo.Title);
duplicateList.Add(item);
}
}
}
}
previousCP = item;
poiProcessed++;
if (poiProcessed % 300 == 0)
{
System.Diagnostics.Debug.WriteLine("Deduplication: " + poiProcessed + " processed of " + totalPOI);
}
}
dupeIdentWatch.Stop();
Log("De-dupe pass took " + dupeIdentWatch.Elapsed.TotalSeconds + " seconds. " + (dupeIdentWatch.Elapsed.TotalMilliseconds / cpList.Count) + "ms per item.");
//remove duplicates from list to apply
foreach (var dupe in duplicateList)
{
cpList.Remove(dupe);
}
Log("Duplicates removed from import:" + duplicateList.Count);
//add updated items (replace duplicates with property changes)
foreach (var updatedItem in updateList)
{
if (!cpList.Contains(updatedItem))
{
cpList.Add(updatedItem);
}
}
Log("Updated items to import:" + updateList.Count);
//populate missing location info from geolocation cache if possible
Stopwatch geoWatch = new Stopwatch();
geoWatch.Start();
PopulateLocationFromGeolocationCache(cpList, coreRefData);
geoWatch.Stop();
Log("Populate Country from Lat/Long took " + geoWatch.Elapsed.TotalSeconds + " seconds. " + (geoWatch.Elapsed.TotalMilliseconds / cpList.Count) + "ms per item.");
//final pass to catch duplicates present in data source, mark additional items as Delisted Duplicate so we have a record for them
var submissionStatusDelistedDupe = coreRefData.SubmissionStatusTypes.First(s => s.ID == 1001); //delisted duplicate
previousCP = null;
//sort current cp list by position again
cpListSortedByPos = cpList.OrderBy(c => c.AddressInfo.Latitude).ThenBy(c => c.AddressInfo.Longitude);
//mark any duplicates in final list as delisted duplicates (submitted to api)
foreach (var cp in cpListSortedByPos)
{
bool isDuplicate = false;
if (previousCP != null)
{
isDuplicate = IsDuplicateLocation(cp, previousCP, false);
if (isDuplicate)
{
cp.SubmissionStatus = submissionStatusDelistedDupe;
cp.SubmissionStatusTypeID = submissionStatusDelistedDupe.ID;
if (previousCP.ID > 0)
{
if (cp.GeneralComments == null) cp.GeneralComments = "";
cp.GeneralComments += " [Duplicate of OCM-" + previousCP.ID + "]";
cp.ParentChargePointID = previousCP.ID;
}
}
}
if (!isDuplicate)
{
previousCP = cp;
}
}
report.Added = cpListSortedByPos.Where(cp => cp.ID == 0).ToList();
report.Updated = cpListSortedByPos.Where(cp => cp.ID > 0).ToList();
report.Duplicates = duplicateList; //TODO: add additional pass of duplicates from above
//determine which POIs in our master list are no longer referenced in the import
report.Delisted = masterList.Where(cp => cp.DataProviderID == report.ProviderDetails.DataProviderID && cp.SubmissionStatus != null && (cp.SubmissionStatus.IsLive == true || cp.SubmissionStatusTypeID == (int)StandardSubmissionStatusTypes.Imported_UnderReview)
&& !cpListSortedByPos.Any(master => master.ID == cp.ID) && !report.Duplicates.Any(master => master.ID == cp.ID)
&& cp.UserComments == null && cp.MediaItems == null).ToList();
//safety check to ensure we're not delisting items just because we have incomplete import data:
if (cpList.Count < 50)// || (report.Delisted.Count > cpList.Count))
{
report.Delisted = new List<ChargePoint>();
}
//determine list of low quality POIs (incomplete address info etc)
report.LowDataQuality = new List<ChargePoint>();
report.LowDataQuality.AddRange(GetLowDataQualityPOIs(report.Added));
report.LowDataQuality.AddRange(GetLowDataQualityPOIs(report.Updated));
Log("Removing " + report.LowDataQuality.Count + " low quality POIs from added/updated");
//remove references in added/updated to any low quality POIs
foreach (var p in report.LowDataQuality)
{
report.Added.Remove(p);
}
foreach (var p in report.LowDataQuality)
{
report.Updated.Remove(p);
}
//remove updates which only change datelaststatusupdate
var updatesToIgnore = new List<ChargePoint>();
foreach (var poi in report.Updated)
{
var origPOI = masterListCopy.FirstOrDefault(p => p.ID == poi.ID);
var updatedPOI = poiManager.PreviewPopulatedPOIFromModel(poi);
var differences = poiManager.CheckDifferences(origPOI, updatedPOI);
differences.RemoveAll(d => d.Context == ".MetadataValues");
differences.RemoveAll(d => d.Context == ".DateLastStatusUpdate");
differences.RemoveAll(d => d.Context == ".UUID");
differences.RemoveAll(d => d.Context == ".DataProvider.DateLastImported");
differences.RemoveAll(d => d.Context == ".IsRecentlyVerified");
differences.RemoveAll(d => d.Context == ".DateLastVerified");
differences.RemoveAll(d => d.Context == ".UserComments");
differences.RemoveAll(d => d.Context == ".MediaItems");
if (!differences.Any())
{
updatesToIgnore.Add(poi);
}
else
{
//differences exist
CompareLogic compareLogic = new CompareLogic();
compareLogic.Config.MaxDifferences = 100;
compareLogic.Config.IgnoreObjectTypes = false;
compareLogic.Config.IgnoreUnknownObjectTypes = true;
compareLogic.Config.CompareChildren = true;
ComparisonResult result = compareLogic.Compare(origPOI, updatedPOI);
var diffReport = new KellermanSoftware.CompareNetObjects.Reports.UserFriendlyReport();
result.Differences.RemoveAll(d => d.PropertyName == ".MetadataValues");
result.Differences.RemoveAll(d => d.PropertyName == ".DateLastStatusUpdate");
result.Differences.RemoveAll(d => d.PropertyName == ".UUID");
result.Differences.RemoveAll(d => d.PropertyName == ".DataProvider.DateLastImported");
result.Differences.RemoveAll(d => d.PropertyName == ".IsRecentlyVerified");
result.Differences.RemoveAll(d => d.PropertyName == ".DateLastVerified");
result.Differences.RemoveAll(d => d.PropertyName == ".UserComments");
result.Differences.RemoveAll(d => d.PropertyName == ".MediaItems");
System.Diagnostics.Debug.WriteLine("Difference:" + diffReport.OutputString(result.Differences));
if (!result.Differences.Any())
{
updatesToIgnore.Add(poi);
}
}
}
foreach (var p in updatesToIgnore)
{
if (report.Unchanged == null) report.Unchanged = new List<ChargePoint>();
report.Unchanged.Add(p);
report.Updated.Remove(p);
}
//TODO: if POi is a duplicate ensure imported data provider reference/URL is included as reference metadata in OCM's version of the POI
stopWatch.Stop();
Log("Deduplicate List took " + stopWatch.Elapsed.TotalSeconds + " seconds");
//return final processed list ready for applying as insert/updates
return cpListSortedByPos.ToList();
}
