Apply Asset Package: float changes to double in enterprise geodatabase
Accepted Solutions
The code above was to show that despite telling the system to create a FLOAT, it created a DOUBLE (implicit upcast). There is a very specific edge case where apply AP does explicitly upcast a FLOAT to a DOUBLE (i.e., it creates the field as a DOUBLE to begin with). You get a runtime warning informing you about this.
If you create a FLOAT field and don't specific scale/precision, then you get a DOUBLE. Tested with the same software as above, so you'll see this behavior even after you upgrade your client.
Edit:
Apply AP has been creating FLOAT fields with the following payload since ~Pro 2.1. Scale/Precision are both 0 because that is how they are stored in FGDB.
<AddField>
<field_name>f_05_03</field_name>
<field_type>FLOAT</field_type>
<field_precision>0</field_precision>
<field_scale>0</field_scale>
<field_length>4</field_length>
<field_alias>f_05_03</field_alias>
<field_is_nullable>True</field_is_nullable>
<field_is_required>False</field_is_required>
</AddField>
I'm testing with Pro 3.3 and SQL Server 16.0.4120.1 (what I have handy).
import arcpy
fields = []
for p in range(1, 10):
for s in range(1, p):
fields.append(
dict(
field_name=f"f_{p:02}_{s:02}",
field_type="FLOAT",
field_alias=f"Precision: {p}, Scale: {s}",
field_precision=p,
field_scale=s,
)
)
for f in fields:
arcpy.AddField_management(
r'<my table>',
**f
)
print(arcpy.GetMessages())The fields are created as numeric(p, s), but ArcGIS treats any precision >6 as a DOUBLE:
Same result here with my version of SQL Server: changes over to Double at the "f_07_01" field (precision > 6).
This clearly illustrates what happens in general when manually creating new fields. I was mainly wondering if anyone out there can confirm "Apply Asset Package" still converts from Float to Double in later versions of Pro / UN schema (3.1+). If I need to upgrade one of our test computers and run through the whole process on it then I will, but there must be many users already on 3.1 right now who have tried this. I'm not totally clear if you're implying the answer is yes (3.1 and higher also convert Float to Double). If that's the case, then I'm confused on how Apply Asset Package is not "explicitly upcasting the field" and yet the field is in fact different in the target EGDB.
The code above was to show that despite telling the system to create a FLOAT, it created a DOUBLE (implicit upcast). There is a very specific edge case where apply AP does explicitly upcast a FLOAT to a DOUBLE (i.e., it creates the field as a DOUBLE to begin with). You get a runtime warning informing you about this.
If you create a FLOAT field and don't specific scale/precision, then you get a DOUBLE. Tested with the same software as above, so you'll see this behavior even after you upgrade your client.
Edit:
Apply AP has been creating FLOAT fields with the following payload since ~Pro 2.1. Scale/Precision are both 0 because that is how they are stored in FGDB.
<AddField>
<field_name>f_05_03</field_name>
<field_type>FLOAT</field_type>
<field_precision>0</field_precision>
<field_scale>0</field_scale>
<field_length>4</field_length>
<field_alias>f_05_03</field_alias>
<field_is_nullable>True</field_is_nullable>
<field_is_required>False</field_is_required>
</AddField>
