Latest Contributions by VinceAngelo

Some indexes use the column being indexed as part of the name. If something international-ish is happening in the field names, then the UTF-16 name (all strings in FGDB are "wide char") could be encoded UTF-8 in the OS (or translated to the local codepage, if it's not UTF-8).  If the Python script is reading these incorrectly, it seems a codepage issue.  Deleting the indexes will not do the FGDB any favors., but so long as the feature count isn't impressive, lack of indexing probably won't doom you. Still, removing files from the OS isn't a recommended procedure. You should probably check the Python script, to see if it is at fault here. - V ... View more

Okay, so I had my replica globalid population corrupted by the failure of arcpy.da.InsertCursor to honor  arcpy.env.preserveGlobalIds, and a search turned this up, so I needed to engineer a solution. If there's only a few features, and a reliable key column, you could use an array to store up key_col(s),globalid list of lists to issue an UPDATE via an arcpy.ArcSDESQLExecute cursor but I have widely disparate row counts in several score of tables, and sending 30M UPDATE commands is not the most efficient way to attack this problem (at 6ms per UPDATE that's 50 hours wasted). What I've settled on is populating parallel tables with a rowid column, the primary key column(s) plus globalid as a VARCHAR(38) column, then running a single update: UPDATE schema.tablename_t1 t SET globalid = vt.globalid FROM ( SELECT t1.objectid,g1.globalid FROM schema.tablename_g1 g1 JOIN schema.tablename_t1 t1 USING (keycolumns) ) vt WHERE t.objectid = vt.objectid Notes: This won't work with versioned tables (I need to populate these tables before enabling versioning, so I'm good) You need to use an arcpy.da.Editor to populate the tables in parallel (two open DA cursors on the same connection won't work without it) I drive the inner virtual table query from the G1 to the T1, because then I don't have to index both tables, just T1 (which I need to index anyway). Don't forget to TRUNCATE the g1 table contents after the UPDATE (I keep the table around for reloading, because I don't want to CREATE/Register/DROP tables and run up the registered rowid sequence indefinitely) - V ... View more

This feels like a networking issue, possibly a flighty router. If you can reproduce the behavior with a simple ODBC connection app then it's not Pro-specific and may help you diagnose the issue. - V ... View more

There are several ways to implement replication, among them: Log shipping from a parent database to a read-only replica, ArcGIS geodatabase replication, and Application-based change detection/management. It's not clear if you are using 1 or 2, and if 2, whether it is one-way or two-way. That said, I've used all three successfully with ArcGIS, including additional variants like standby replicas. You can also combine these (e.g., both 1 & 2, or 3 & 1 where the map services are hosted from the read-only replica). - V ... View more

It's not clear what your use case is. Generally, I'd be wary of unnecessary 2-way replicas. Editing a single parent geodatabase and sharing it with two parent-to-child one-way replicas is the simplest (and most common) data model available, far easier than two two-way replicas. ArcGIS 11.3 is somewhat elderly at this point, so if you're only starting now, I'd recommend using something more recent.  Maybe not the brand new stuff, but certainly the most recent patch to the stable release before that (which in this case would be ArcGIS 12.0/Pro 3.6.4, or ArcGIS 11.5/Pro 3.5.7, depending on your view of .0 releases). - V ... View more

Ah,... yes.  Well, this is more a religious issue than a technical one. I am not among those who value precision because it is possible, rather than needed. I do see significant performance benefits from my somewhat heretical techniques, though not as much in native RDBMS geometry storage, because non-Esri geometry' doesn't see the same storage reduction found in file geodatabase or ST_GEOMETRY. If your data is only accurate to 1-inch, then 1/0.01' or 1/0.005' could be reasonable XY scalefactors, The general tools don't make it easy to be consistent in non-default coordinate references, but I've found that using a "Transfer" FeatureDataset with the desired coordref constants as an intermediary for creation, then immediately dragging the newly created FCs out of that FDS to be effective.  I usually script my creation utilities, which allows for more flexibility in SpatialReference definition and application, but the FDS technique works quite well for ad-hoc geoprocessing outputs. - V ... View more

Okay, so it's a bit more complicated to drive from the joiner table...  Because some of the tables have compound keys, I have to isolate only changed rows by joining the joiner back to the table in the subquery, returning table objectids: WHERE objectid in ( SELECT t.objectid FROM schema.table_j1 j JOIN schema.table_t1 t USING (keycol1,...,keycolN) )   And if it's versioned, then you have to join to the versioned view: WHERE objectid in ( SELECT t.objectid FROM schema.table_j1 j JOIN schema.table_t1_evw t USING (keycol1,...,keycolN) ) - V ... View more

I've been using PostgreSQL for so long I've forgotten how to spell SQL Server. My joiner tables are regular tables with the same base name as the primary table (business table "source.table_name_t1", digest table "source.table_name_h1", joiner table "source.table_name_j1"). They all share the primary key column(s), but t1 has the row values, h1 has the row digest, and j1 has the  u_or_d column. I name the oid columns differently so the joined result is unambiguous. INSERTs are executed immediately by an arcpy.da.InsertCursor.  There really isn't likely to be much performance gain by writing to an extra table, then appending from that, though I suppose it depends on the number of indexes in the primary table.  I have a further complication of needing to do this to a replica parent geodatabase, and using "INSERT INTO ... SELECT ..." with versioning would add unnecessary complexity.  Performance-wise, the query from a source DBMS is the long pole in my tent, and SHA-1 hashing is somewhat expensive, especially when less than 1% of each table is modified, so there's large input I/O and compute costs, actually writing new rows is somewhat rare. Generally speaking, I don't permit empty strings to kludge around NOT NULL constraints, but I'm not always able to enforce that. For purposes of hashing, I'd recommend treating an empty field in a CSV as equivalent to a NULL. I have a funky table or two where I have to backtrack to enforce NOT NULL on strings by replacing None with ''  before INSERT, but doing it that way makes me feel dirty. FWIW, I make extensive use of function dictionaries, so my encoder functions are placed in a dictionary by target datatype, and then they can massage oddities without harming the data stream. The code below handles empty string and NULL differently, (a single 0 byte vs a single ETX control character byte) def pack_str(v,n): if (v == None): return(b'\x00') ba = bytearray() if (type(v) == datetime.datetime): if VERBOSE_ENABLED: print(" {:8d}: Expected str, got {:s}!".format(n,str(type))) v = v.strftime(isoFmt) elif (type(v) != str): if VERBOSE_ENABLED: print(" {:8d}: Expected str, got {:s}!".format(n,str(type))) v = str(v) vt = v.rstrip(' ').encode('UTF-8') ba.extend(vt) ba.extend(b'\x03') # ASCII End-Of-Text (ETX) return ba def pack_i2(v,n): if (v == None): return(b'\x00') return struct.pack('h',v) def pack_i4(v,n): if (v == None): return(b'\x00') return struct.pack('i',v) ... packer_by_type = { # Lookup via native 'udt' value 'bool' : pack_i2, 'bpchar' : pack_str, 'bytea' : None, # BLOBs manifest as memoryview (unsed) 'char' : pack_str, 'date' : pack_date, 'float4' : pack_f8, 'float8' : pack_f8, 'geography' : pack_geom, 'geometry' : pack_geom, 'int2' : pack_i2, 'int4' : pack_i4, 'int8' : pack_i8, 'numeric' : pack_f8, 'timestamp' : pack_date, 'text' : pack_str, 'uuid' : pack_str, 'varchar' : pack_str, } # Note: When object is instantiated, the packing function is added # to a "packers" list ... def getHash(self,qRow): self.rowCount += 1 ba1 = bytearray() for i,packer in enumerate(self.packers,start=self.keyCount): ba1 += packer(qRow[i],self.rowCount) return hashlib.new('sha1',ba1,usedforsecurity=False).hexdigest()     ... View more

I've done this several times now, once with hundreds of thousands of "changed" records twice daily (where only 10% are actually new), and once with 28 million records in each delivery (full dataset). What I settled on was storing SHA-1 hashes of the data rows in a parallel table in the database.  The loader walks the tables, loading all the digest records first, then hashing the input stream with the same algorithm.  If the key column(s) are not present in the dictionary of hashes, it's an INSERT; if the key exists, and the SHA-1 the same, that's a No-Op, and I delete the key from the dictionary; if the key exists, but the SHA-1 is different, I tag it as an update, then also delete the dictionary key. Once I've processed all rows, any keys that are left are my delete candidates.  Initially, I just walked the UPDATE and DELETE lists, but that sometimes made for hundreds of thousands of UpdateCursor calls, and running a full scan on 20+ million rows to change 200K wasn't cost-effective.  So I tweaked the initial solution to write records to a "joiner" table, with the key and 'U' or 'D', then ran one UpdateCursor on the table, with a "WHERE keycol in (SELECT keycol FROM joiner_tab)" query.  Then I could join the joiner to the primary table and the digest table to manifest the changes in two DA cursor calls (one for each table). Using this technique I preserve the map service integrity during modifications (one COMMIT per table), without the outage caused by a TRUNCATE/re-load.  Yeah, there might be occasional digest collisions that produce a false positive, but with similar data, collisions are exceedingly rare. This methodology is also highly reliable -- One of these systems has been running with zero maintenance (beyond database and ArcGIS Pro upgrades) for over five years. In theory, the API for Python could do this through the Portal, but in practice, using an UpdateCursor in ArcPy is literally a thousand times faster (680ms v. 12 minutes for 39K rows) than having to process each row as a series of XML web requests through the Portal and on to the AGS host, then finally to the database. I did this all with plain vanilla ArcPy. The only tricky part is dealing with values coming in as database rows and as ASCII text, since type preservation can be a challenge.   I've also created parallel tables to insert all rows into, letting the database manage the conditional updates as a trigger, but that's better suited to a fixed number of records (states, countries, counties, zip codes,...). Good luck! - V ... View more

I'm not a big fan of coordsys offsets with large numbers of significant digits.   0.003281 would be an *awful* precision to use.  I would choose 0.001 or 0.002, since the conversion to integer is "times 1000" or "times 500" (and not "times 304.78512648582749161840902163974". Precision to a thousandth of a foot is gratuitous overkill for all but precision amoeba farming, but it's a nice round number. I try to avoid the default values, because they were chosen to maximize precision, at the cost of compressibility, and I'd rather transfer a BLOB with 278 bytes than one with 40K (your mileage may vary, but you should take your most complex geometries and try them at different scalefactors, to determine what the exact compression loss might be). - V ... View more

Copy/delete/rename is far from the most efficient methodology here. I have tables with tens of millions of rows in them; doing this would cost me hours and could result in tables with columns slightly different from the datatypes I originally declared them to be. I pretty much avoid Copy for this reason. - V ... View more

Brian -- There are potential issues to managing records in a registered table in this manner. The easiest solution is to just leave the feature class unregistered, and access it as a Query Layer.  But make sure you recluster the table, since doing this sort of "load, then update geometry" process thoroughly fragments the table.  Truncate and reload is not a very efficient process, and leaves any service looking at the table during that transition somewhat bereft. I prefer to do change detection, and only INSERT new records, UPDATE changed records, and DELETE removed ones. I'm doing this daily with tens of millions of rows and hundreds of thousands of potential changes, in just minutes over the time it takes to read the new records. - V ... View more

Alicia -- The RDBMS (or file geodatabase handler) is not aware of feature datasets, so even though you see it "in" a feature dataset, it isn't really, and you can't have duplicate-named feature classes, so you can't do a straight copy. I just drag the feature class out of the feature dataset to be standalone, then drag it back in to a different feature dataset, but I just tested, and it's possible to just drag from dataset1 to dataset2 without copy.  So long as the Spatial Reference is the same, it's no big deal. If the spatial references are different, then you have to rename to "_old" then use FeatureClassToFeatureClass to change the SpatialReference. But you might want to re-evaluate whether you need to have feature classes in a feature dataset at all.  If they're not being used for "feature-dataset-ish" tasks like topologies, then having more standalone feature classes would reduce the overhead of locking all the feature dataset members whenever a lock is placed on any feature class in the dataset. - V ... View more

Barbara -- There's probably a two-billion-ish limit on the number of concurrent states, but if you were to attempt it, your server would likely fail from old age before it completed. Keep in mind that a "version" is a named reference to a state in the state tree, and that you're unlikely to have 2^31-1 versions, because each edit is usually its own state, and there's usually more than one edit in any edit session. I doubt you'd be so lucky that some posts would succeed in such a situation, since it's a multi-pass process.  If you want to find effective limits, I suggest you conduct a benchmark of sorts, with multiple edits per version, with increments of 500 versions at a time. You should be prepared for this to take an inordinate amount of time, and for it to eventually corrupt your instance, so I wouldn't suggest you do this on a production system. - Vince ... View more

Paige - Cool.  But don't forget to check if there are other, multi-part types in the table: SELECT ST_GeometryType(geomcol), count(*) as nfeats FROM owner.geomtable GROUP BY ST_GeometryType(geomcol) ORDER BY 2 DESC Note that the database name isn't required, because PostgreSQL doesn't support cross-database queries. - V ... View more