[Feature] Log Browsing History over multiple sessions. about rtv

[Brobin]

One issue I think I will run into is having a very large file (thousands of links). Perhaps links older than 1 month are not saved

[Brobin]

Also, I'm not too sure about log file standards. Would it be okay to put it in .config/rtv/history.log?

[michael-lazar]

From what I can find it either needs to go in XDG_CACHE_HOME or XDG_DATA_HOME. It's unclear to me which one is correct, but I would lean towards CACHE because it's non-essential data. I would also name the file history.txt because it's not a standard log in the sense that it will be consumed and overwritten by the program.

Using timestamps for each link sounds like a lot of overhead, it might be better to just save the most recent N links.

[Brobin]

To save the most recent N links, we would have to use some sort of ordered list. A set is unordered so restricting the size of it when writing would not guarantee that you would get the last N links, just N random links.

Would you be opposed to refactoring to use a list instead? That way we could just slice the end of the list for the output file.

Also, is 1000 a reasonable number of links to save?

[michael-lazar]

I used a set because testing for membership is O(1) as opposed to lists which are O(N). I don't think going back to a list is a good idea. Also think about how you're going to write to this file. If you only write when the program is quit via q, you will be losing history on ctrl-c.

Here's one approach that uses file appends to avoid having to overwrite the whole file after each time. There may be other approaches as well.

1. Read the file in as a list.
2. If the list length is over N, overwrite the file with only the most recent N links.
3. Convert the list to a set
4. When a new link is clicked, add it to the set AND append the link directly to the end of the file.

As for the number of links, I think 100-200 would probably be enough.

[Brobin]

Good point about the complexity. I like this approach. It has lot of extra file operations, but I guess that's necessary to not lose history if you quit with ctrl+c.

[noahmorrison]

As for the ctrl+c problem, why not use atexit?

import atexit
import time

@atexit.register
def save():
    print('saving')

if __name__ == '__main__':
    time.sleep(2)
    print('done')

run this, no matter what (ctrl+c, or normal exit), "saving" will always be printed.

[Brobin]

Didn't know that module existed! Sounds good to me. If we use that, I think it would be best to do the file I/O at exit instead of when you visit each link. It would keep the code cleaner.

[Brobin]

I've got an implementation working. @michael-lazar, it is very similar to how you described it, except it uses the a@atexit decorator to do one file operation at exit instead of loading and writing to the file all of the time. See #95

[michael-lazar]

Merged in #95