Gerd, I have commited your patch plus the header length implementation. The int for the record layout should give the format of the records not the header. So I don't see a reason to change the record layout id if I just want to add a field to the header and therefore I need the header length. WanMil
Hi Gerd,
the length field is required to be able to add more fields to the header in the future. Example (please don't nail me with the given length of the fields - it's just an example):
Header: Format type: String: BND Timestamp: long | Header length: int: 8+4+10+4 = 26 bytes | Record format: String: RAW (8 bytes) | Record layout: int: 0x01 (4 bytes) | mkgmap release: String: 2248 (10 bytes) | Additional id: int: 0x45 (4 bytes) ... Records ...
Now think about three different reader versions all trying to read that file:
Reader "old" does not know about the new additional id. But it can read until the mkgmap release and can skip the 4 bytes fo the additional id (due to the known header length).
Reader "current" expects exactly all fields until the additional id. It reads 26 bytes and finishes.
Reader "future" expects an additional field "Java version". But after reading the additional id it can detect that the header does not contain this new field "Java version". Therefore it can assume a default value for that.
If you don't use the header length you won't have the chance to add anything to the header. (You've had the same problem with the legacy format...)
WanMil
Hi WanMil,
I don't see a need for the length field in the header, and I don't see how I could use it (I found no way to calculate how many bytes were removed from a stream)
So I've coded this: Format type: String: BND Timestamp: long | Record format: String: (RAW, QUADTREE) | Record layout: int | mkgmap release: String ... Records ...
Attached is the patch that implements this.
Gerd
Date: Sun, 11 Mar 2012 21:02:56 +0100 From: wmgcnfg@web.de To: mkgmap-dev@lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] [Patch] Error in BoundaryUtil
Hi WanMil,
Date: Sun, 11 Mar 2012 20:08:54 +0100 From: wmgcnfg@web.de To: mkgmap-dev@lists.mkgmap.org.uk Subject: Re: [mkgmap-dev] [Patch] Error in BoundaryUtil
i Gerd,
can you please explain in short what's the difference between the three formats (my thinking): * Legacy (original - int coords) * Raw (float coords) * Quadtree (admin level areas merged to the quadtree format)
yes, correct
loadQuadTree (Quadtree format?) => loadQuadTreeFromStream (Quadtree format?) => readStreamRawFormat (Raw format?) => readArea (Legacy
format)
At least it is irritating to me if there is a method
readStreamRawFormat
which reads raw and/or legacy format.
I agree. I'll change that so that we have two methods readStreamLegacyFormat and readStreamRawFormat.
Great! Maybe it makes also sense to put the detection of the format into one method? In the moment there are a lot of rel.endsWith(...) checks in the code.
I would also propose a change to the header of the new raw and
quadtree
format. In the legacy format the first String in the file gives the release of mkgmap. You added a "_raw" or "_quadtree" (?) to mark the new formats. This is unclean but required to support the legacy format. But legacy format support can be dropped in the future (I think at least in the end of 2012). So to be open for future other changes it would be good to have a kind of format identification. This could be added after the creation time. Having this format identification future changes to the format could be easily and much cleaner added.
Yes, my problem with the legacy format was that it did not contain such a format identifier. I've worked a long time with programs on IBMs mainframe system (OS/390, z/OS). They typically use an eye catcher to identify the record type and a numeric field that identifies the record layout version. In some programs, they have also a length field. Old programs can read newer structures because the newer fields are only added at the end of a structure, and the length field allows to skip them. I think this is a good idea, but we probably don't have to support all kinds of versions in programs and files.
So, here is my suggestion: instead of adding _raw or _quadtree to the 1st string I replace it with BND. Legacy format is identified because it doesn't contain this string. A new String field fileFormat after the timestamp identifies the file format, either RAW or QUADTREE. One more int field recId identifies the record layouts used (0 for this first version). Whenever a new format is needed, we change fileFormat, if only a new field is needed in e.g. the AREA section, we increase recId. All methods that read the file verify that they are able to process the recId or they stop with an FormatException . OK?
Sounds good. But I propose two additional things: 1. a header size or id (otherwise we will have the same problem again to add an additional information to the header) 2. Keep an mkgmap release identifier.
So the new format would look like: Format type: String: BND Header size: int | Record format: String: (RAW, QUADTREE) | Record layout: int | Timestamp: long | mkgmap release: String ... Records ...
WanMil
Gerd
WanMil
Hi WanMil,
hmm, we have two errors here. I can confirm that my patch
doesn't solve
all problems, it just allows again to read the old *.bnd format and the new quadtree format, but mkgmap fails to read the intermediate format produced by the preparers 1st pass.
> Date: Sun, 11 Mar 2012 18:25:21 +0100 > From: wmgcnfg@web.de > To: mkgmap-dev@lists.mkgmap.org.uk > Subject: Re: [mkgmap-dev] [Patch] Error in BoundaryUtil > > Hi Gerd, > > without your patch preparing boundaries works.
No, it doesn't. It stops (crashes) without any error message when trying to read the *.bnd files produced in the 1st pass. You can verify this by looking at the header of the new *.bnd files, they have the suffix _raw, not _quadtree.
The new patch fixes this, but not the missing error detection.
I think we need a try/catch block somewhere, either in Preparer or BoundaryPreparer. Will you take care of this?
Gerd
> Using your patch I get the following exception. > > java.util.concurrent.ExecutionException: > java.lang.IllegalArgumentException: Ill > egal Capacity: -572471432 > at java.util.concurrent.FutureTask$Sync.innerGet(Unknown Source) > at java.util.concurrent.FutureTask.get(Unknown Source) > at > uk.me.parabola.mkgmap.main.Preparer.runPreparer(Preparer.java:92) > at uk.me.parabola.mkgmap.main.Main.endOptions(Main.java:349) > at > uk.me.parabola.mkgmap.CommandArgsReader.readArgs(CommandArgsReader.ja > va:126) > at uk.me.parabola.mkgmap.main.Main.main(Main.java:114) > Caused by: java.lang.IllegalArgumentException: Illegal Capacity: -572471432 > at java.util.ArrayList.<init>(Unknown Source) > at > uk.me.parabola.mkgmap.reader.osm.boundary.BoundaryUtil.readStreamRawF > ormat(BoundaryUtil.java:282) > at > uk.me.parabola.mkgmap.reader.osm.boundary.BoundaryUtil.loadQuadTreeFr > omStream(BoundaryUtil.java:459) > at > uk.me.parabola.mkgmap.reader.osm.boundary.BoundaryUtil.loadQuadTree(B > oundaryUtil.java:145) > at > uk.me.parabola.mkgmap.reader.osm.boundary.BoundaryUtil.loadQuadTree(B > oundaryUtil.java:120) > at > uk.me.parabola.mkgmap.reader.osm.boundary.BoundaryPreparer$QuadTreeWo > rker.call(BoundaryPreparer.java:236) > at > uk.me.parabola.mkgmap.reader.osm.boundary.BoundaryPreparer$QuadTreeWo > rker.call(BoundaryPreparer.java:221) > at java.util.concurrent.FutureTask$Sync.innerRun(Unknown Source) > at java.util.concurrent.FutureTask.run(Unknown Source) > at java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown > Source) > at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown > Source) > at java.lang.Thread.run(Unknown Source) > > My mkgmap call is: > java -jar mkgmap.jar --max-jobs=4 > --createboundsfile=africa-boundaries.osm.gz --bounds=bounds/africa *.osm > > WanMil > > > Hi WanMil, > > > > sorry, one of my last patches for the performance branch corrupted > > BoundaryUtil. > > It was no longer able to read the legacy *.bnd format (also created in the > > preparer) :-( > > > > Attached is the patch. > > > > Gerd > > > > http://gis.19327.n5.nabble.com/file/n5554370/BoundaryUtil.java.patch > > BoundaryUtil.java.patch > > > > -- > > View this message in context:
http://gis.19327.n5.nabble.com/Patch-Error-in-BoundaryUtil-tp5554370p5554370...
> > Sent from the Mkgmap Development mailing list archive at Nabble.com. > > _______________________________________________ > > mkgmap-dev mailing list > > mkgmap-dev@lists.mkgmap.org.uk > > http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev > > _______________________________________________ > mkgmap-dev mailing list > mkgmap-dev@lists.mkgmap.org.uk > http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________ mkgmap-dev mailing list mkgmap-dev@lists.mkgmap.org.uk http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev