JabRef 4.3.1
Linux 4.18.14-arch1-1-arch amd64
Java 1.8.0_192

Related TSE question: https://tex.stackexchange.com/questions/436684/correct-arxiv-hyperlinks-with-biblatex-and-mendeley

Steps to reproduce:
Random example: https://arxiv.org/abs/1810.09995

1. Use biblatex mode and biblatex format
2. Import a publication by its arxiv URL.
   The problem: unlike with DOI (e.g. it easily parses https://doi.org/10.1137/16M1104986 but not doi.org/10.1137/16M1104986), arxiv cannot discard https://arxiv.org/abs/ URL prefix
3. Import a publication by its ID.
   Result:

@Article{marcheggiani2018dgc,
  author      = {Diego Marcheggiani and Laura Perez-Beltrachini},
  title       = {Deep Graph Convolutional Encoders for Structured Data to Text Generation},
  date        = {2018-10-23},
  eprint      = {http://arxiv.org/abs/1810.09995v1},
  eprintclass = {cs.CL},
  eprinttype  = {arXiv},
  abstract    = {...},
  file        = {:http\://arxiv.org/pdf/1810.09995v1:PDF},
  keywords    = {cs.CL},
}

4. Cite this using biblatex, and it reads

D. Marcheggiani and L. Perez-Beltrachini. “Deep Graph Convolutional En-
coders for Structured Data to Text Generation”. In: (Oct. ...). arXiv:
https://arxiv.org/abs/1810.09995 [cs.CL].

The problem: biblatex generates wrong link because of wrongly formatted eprint: https://arxiv.org/abs/http://arxiv.org/abs/1810.09995v1

Summary

When importing an arxiv preprint, JR only accepts ID, but it puts the full URL into the eprint field, while it should contain only the ID.

Suggestions

• unify URL parsing: protocol and website name should not matter
• write only ID to the eprint field