1

I'm hoping someone out there might be able to help or suggest some tests to try and resolve some very odd QGIS behaviour of late...

Our core GIS team of 4 users has been testing v2.10 prior to roll-out for the remainder of the organization.

In the past couple of weeks we have had some very odd shapefile corruption things happening and are yet to isolate or identify a potential cause...

These have included:

  1. Editing one line (using the scissor tool to split into two records) and having the resulting records attributes completely replaced by attributes from another geometry no-where near the edited row (spatially or in the attribute dbf)
  2. Line geometries completely disappear while their attributes get associated with an unrelated geometry
  3. Line geometry get partially deleted and replaced with a vertice in infinity (line visually extends beyond the dataset and no matter how far zoomed out you never see the end)

The above is happening in a polyline layer, that we did discover has some multi-line geometries..? Perhaps this is impacting... but never saw this issue with v2.8 and the same layer. The layer has undergone MANY save-as's in the past week and many verify geometry checks as we thought invalid geometry may have been the cause...

In another separate layer yesterday, to a separate user (me) i created a layer with approx 30 polygons... adding, deleting some and doing some 'avoid intersection' to remove overlaps. Layer looks fine in qgis, saved and also a save-as to a completely new file but when i load into ArcGIS online (and later is ArcMap as a test) some of the overlaps and deleted objects are still there. Close and reopen qgis, reopen the file and they are definitely not.

The project was created in v2.8 that i was editing in (but had been saved in 2.10).

Does anyone have any ideas/suggestions.

We do still have 2.8 installed as well as 2.10? We have never encountered issues like this when testing other new qgis versions while still running old ones?

  • I believe that there are are some open tickets about phantom features. Theoretically you can delete features from a shapefile by marking the row in the .dbf file as deleted but without wiping our their row from the .shp and .shx files physically. This is fast because large rewrites are avoided but seems to be error prone. But read more from the QGIS tickets, I may remember wrong. – user30184 Aug 21 '15 at 8:57
  • 1
    For suspected bugs, probably best to (1) do a search on the issue tracker to see if it's been reported, and please report it if not, and (2) check the osgeo forums. – Simbamangu Aug 21 '15 at 9:00
  • Thanks. I will check both forums ans issue tracker. I did find some references before i posted about shapefiles getting corrupted with the number of rows in database and geometry out of sync... but these seemed to be resolved by a save-as which we have done and are continuing to see random errors. We also haven't seen any "_packed" table names which these others refer too... – Jamie Aug 21 '15 at 9:59
  • Re. Logging an issue, i was just hoping to determine a little more about when/why it happens to provide some kind of help to trace the cause.... so far yet to identify any common theme other than the layer has been edited! As many of the corrupt records are not at all related to what was being edited at the time, we have sometimes not noticed until hours later.... – Jamie Aug 21 '15 at 10:02
  • Similar problems here. I’m a single user digitising and editing property boundaries and adding/editing attributes data as a shape file. QGIS 2.10.1 (Upgraded from 2.8 via osgeo4w) will sometimes randomly "lose" polygons and then assign data to the wrong polygons on save. I’ve had to redo work several times over. Some rasters (ecw) which imported fine in 2.8 are now several 100 meters out while neighbouring raster squares are correct. This behaviour is unpredictable so difficult to diagnose. Never had these problems with 2.8. – goatwillow Aug 21 '15 at 11:59
1

We'va had similar issues. The reason was we had several persons editing the same shapefile on the server at the same time, which destroyed the file. Columns and rows were shifted in the .dbf. Try to avoid simultaneous access.

ArcGIS sometimes doesn't handle shapefiles correctly if the ID column is missing or empty. You should check this.

  • Thanks Flo.... we too have seen this happen if more than 1 persons inadvertently edits the file (or even the same person with it duplicated in a project).. – Jamie Aug 21 '15 at 10:06
  • Oops... but this was not the case in either of these two files - definitely only one person editing or even accessing the file at the time the corruptions occurred. – Jamie Aug 21 '15 at 10:07
  • 1
    Editing the same shapefile is not going to be a a good idea, they are not designed for that kind of thing. Using a database or sqlite is the best for that setup. – Nathan W Aug 21 '15 at 10:33
  • The ID column can't be missing or empty -- It doesn't really exist (it's just the record number, manifested as a field). If a column named "ID" exists, ArcGIS will rename it, so the virtual ID can be present. – Vince Apr 27 '16 at 21:26
  • I guess that's why Arcgis created lock files, to avoid same-time editing – gisnside Aug 27 '16 at 16:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.