Please make the parser more robust

[cskyan]

I encountered some ParseError exceptions when I tried to parse the ntriples files. Some of them are quite easy to be fixed during the runtime such as empty lines, codec issues, etc. I hope that the parser could pre-process the files and deal with these problems or ignore the invalid records. At least, we need to know which lines in our data file have problems. Because we cannot make sure that the downloaded files strictly follow the standard format. If the package just raises the exception without correcting it, it will take more time to parse the whole file. Maybe the impact of neglected data could be accepted when we are processing a large data set.
In my case, I directly modify this line of code. I insert a continue code here to let the program proceed. Otherwise, I cannot get the remaining data when I encounter a ParseError. I know that it is not a good way to skip this exception but it is the fastest way to continue my project. Hope that this suggestion would be accepted.

[joernhees]

hmm, i'm against changing the default behavior (it's correct to raise an error if the input format is broken)... but maybe providing an ignore_errors flag via parse() would be a nice feature here...

[cskyan]

Sure, ignore_errors would be a helpful feature in parsing the raw data. And the program should also put the ignored records into the log. Another expected feature is that it could incrementally parse the files. Say if I have some invalid records in my raw data, after the first parsing I can correct them and parse these revised records. It would save lots of time if this feature could also be implemented.

[gromgull]

This is a good feature for the #283 clean-up, then there will be an error-callback.

[mehak16163]

Hi,
I have attempted to solve this issue. Please go through my PR #1080
Please suggest any changes to be made.

[karish-grover]

I have been working on this on this issue for quite some time now. I have made incorporated a few parsing errors, but I was wondering if you could explain in detail, the exact errors that you want us to handle so that I can keep those in mind before sending out a PR.