Return error if argument to parse() is a path to a non-existing file

[manulera]

Hello Bjorn,

First of all, thanks for making pydna. I think it's a great library, and it is easy to follow the logic of functions and classes.

I am using it to make a web API, if you want to have a look: ShareYourCloning. At some point I would love to discuss with you about the project.

I just had some bug in my code that took me a while to find. I had mistyped the name of a file, so the parser would return an empty sequence. I thought there was a problem with the file. The thing is that in parsers.py the function parse will not return an error if one passes a path that does not exist. As you say, it is a greedy function and it tries to read the path as a sequence.

I have looked a bit on how to check if a string could be a path, and I didn't find any simple one-liner. I guess the easiest fix to make it back-compatible would be add an extra optional argument input_is_path or something like that default None:

• None, do what it used to do
• True try to read the file
• False try to parse

I guess the case where one would pass both paths and sequences is not very realistic. Let me know if that would be a possible fix and I can make a pull request with the fix and a test.

[BjornFJohansson]

Hi, and thanks for this input. As you say, parse is a very greedy function, perhaps too much so.
What do you think of splitting it up into several parse_from_x functions?

for example:

• parse_from_string()
• parse_from_path()

This way, the code might be easier to read.

[manulera]

Sounds good, not sure what the best practice is in a case like this. For back-compatibility I guess it's good to keep the function parse, and call parse_from_string or parse_from_path inside parse?

I think it could be entirely removed, but then it would not be back-compatible.

[BjornFJohansson]

I would like to keep parse for now, I have lots of old code depending on it.
It could be something that calls parse_from_string and then parse_from_path.

[BjornFJohansson]

@manulera I think parse_from_string could be replaces by simply using Dseqrecord("string")?