You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The high-level syntax of a configuration property value depends on a couple of factors right now:
The converter being used
The type of configuration source being used
For example, if I have a properties file, and I have a configuration property value foo,bar, this value has different syntax requirements and limitations depending on whether this value is a list or a simple string. Furthermore if I have a YAML-backed configuration source, which already natively supports lists, the source has to "collapse" the value back into a string list again, which makes little sense.
The fact is that our converter rules presently assume that the backing value is always a simple string in a properties-like file, and this is severely limiting for users.
We should impose syntax rules for aggregate types on a per-ConfigSource basis. The ConfigSource should be able to return either an unconverted string, or (if the config source supports it) a native numeric or list value, without having to go through superfluous conversion. This might mean that ConfigSource implementations need access to the target type and/or converter, but this might be a logical consequence of #43 and its successors.
Describe the bug
The high-level syntax of a configuration property value depends on a couple of factors right now:
For example, if I have a properties file, and I have a configuration property value
foo,bar
, this value has different syntax requirements and limitations depending on whether this value is a list or a simple string. Furthermore if I have a YAML-backed configuration source, which already natively supports lists, the source has to "collapse" the value back into a string list again, which makes little sense.The fact is that our converter rules presently assume that the backing value is always a simple string in a properties-like file, and this is severely limiting for users.
We should impose syntax rules for aggregate types on a per-ConfigSource basis. The ConfigSource should be able to return either an unconverted string, or (if the config source supports it) a native numeric or list value, without having to go through superfluous conversion. This might mean that
ConfigSource
implementations need access to the target type and/or converter, but this might be a logical consequence of #43 and its successors.This relates strongly to #545.
The text was updated successfully, but these errors were encountered: