On Thursday, 5 April 2018 at 17:06:04 UTC, rumbu wrote:
Is there a standard way to handle errors in a chain of range
transformations?
Let's say I want to read some comma separated numbers from a
file.
auto myArray = file.byLine().splitter().map!(to!int).array();
Now, besides fatal errors (like I/O), let's suppose I want to
handle some errors in a silent way:
- skip unicode decoding errors;
- assume blank records with 0;
- skip the line entirely if the conversion to int is not
possible;
I can catch UTFException, ConvException or
ConvOverflowException for the entire syntax chain but there are
some disadvantages:
- I don't know exactly which of the chain members thrown the
exception;
- I cannot skip/handle the error and continue;
The only solution I thought about is to break the nice chain
syntax and handle errors on each of the chain members, but I
wonder if there is probably another way.
Thanks.
You could use predicates:
```
string list = "3, 5, 1, , not a number, 100";
int[] numbers = list.split(",").filter!((entry) {
if (!isNumeric(entry.strip))
return false;
else // ... even more cases?
return true;
})
.map!((e) => e.strip.to!int).array;
assert(numbers == [3, 5, 1, 100]);
```
https://run.dlang.io/gist/413282d9726dbac137bf5f35033a8eea
