Fixing dpath to work for any type of leaf nodes

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.

[nayyarv]

fromdecimalimportDecimalts= {
"foo" :{
"dude": Decimal(1.5)
}
}
>>>dpath.util.get(ts, "foo/dude")
TypeError: objectof type 'decimal.Decimal'hasnolen()

Given that leaf checks for a finite number of types any kind of odd type such as a function or custom object will simply not work in this dictionary. I've suggested an approach based on duck-typing, in which I check if the object has a __getitem__ instead of checking particular type (though I do that too for the case of string)

I have a couple of places I wasn't sure about and would love some opinions.

1. When you have an object with a __getitem__ but no items or len the kvs method breaks. I've kind of just returned an empty iterator here, but it should be accessible via keys and doesn't really need to have a walk run on it if there is no glob for that field. I currently just treat it as a leaf, but maybe this might be addressed with some kind of warning?
2. I also was worried about catching a generic Exception in thing[0] since it could easily be either an indication the object is mappable, but just uses a custom Exception and might have a len/items implemented. I've let it stand right now, but am open to change.

[nayyarv]

Another point of concern I had was the fact that we could have an object that implemenets getitem but not items or len.

Arguably since obj["key"] works, doing dpath.utils.get(obj, "key") should work, but it won't since it has to do a walk despite having no globs in the path.

I've left this unfixed, but added an expected fail test in the meantime. I think it's fair to require some iteration protocol when there's a glob for the current part, but otherwise it should be able to work.

[calebcase]

Split this into a new issue: #127

[nayyarv]

Closing this since there is another PR fixing the base issue.