（PRAW）从评论中获取一个值，然后用另一列的数据作为答复，该数据与第一条数据在同一行

Question

Here's the CSV code.

import praw
import time
import csv
import codecs
import re
from collections import defaultdict 

def read_csv():
with open('CPU-Bench.csv', 'r') as f:
    reader = csv.reader(f)
    for row in reader:
        import pandas as pd
        df = pd.read_csv('CPU-Bench.csv')
        saved_column = df.URL #you can also use df['column_name']
    print saved_column

saved_column = df.URL #you can also use df['column_name']


model_url_dict = read_csv()

The CSV file is kind of like this:

Type,Part Number,Brand,Model,Rank,Benchmark,Samples,URL
CPU,BX80671I76950X,Intel,Core i7-6950X,1,117,25,http://cpu.userbenchmark.com/Intel-Core-i7-6950X/Rating/3604
CPU,BX80671I76900K,Intel,Core i7-6900K,2,112,28,http://cpu.userbenchmark.com/Intel-Core-i7-6900K/Rating/3605
CPU,BX80671I76850K,Intel,Core i7-6850K,3,102,55,http://cpu.userbenchmark.com/Intel-Core-i7-6850K/Rating/3606
CPU,BX80648I75960X,Intel,Core i7-5960X,4,102,1651,http://cpu.userbenchmark.com/Intel-Core-i7-5960X/Rating/2580
CPU,BX80662I76700K,Intel,Core i7-6700K,5,98.5,21550,http://cpu.userbenchmark.com/Intel-Core-i7-6700K/Rating/3502
CPU,BX80671I76800K,Intel,Core i7-6800K,6,97,103,http://cpu.userbenchmark.com/Intel-Core-i7-6800K/Rating/3607

I'd like to make it so that if a user says "!benchmark i7 6950x", or "!benchmark i7-6950x", "!benchmark Core i7-6950x", or "!benchmark Intel Core i7-6950x", it looks for the string after !benchmark (in this case, i7 6950x), and, finds that, looks at the URL column, and replies with "Here's some benchmarks for "

But, in place of the is the chosen CPU (again, in this case, i7 6950x)

And in place of url_column is the URL of that CPU (in this case, http://cpu.userbenchmark.com/Intel-Core-i7-6950X/Rating/3604 )

Sorry if that's confusing, but, how do I do this?

Answer 1

You definetely need to setup reddit agent, authenticate it as reddit application in settings and all that other stuff that you do for normal bot. Oh and also it would be nice to parse dataframe to dictionary for easy lookup (since you will only do it when bot starts and then have really fast time to look up).

So we will need dictionary like this {'Intel Core i7-6950x':'http://cpu.userbenchmark.com/Intel-Core-i7-6950X/Rating/3604', ...} , which is pretty trivial considering you already have that read.

We also need a way to parse what user actually wants to benchmark. So parse_models_from_comment(comment) that would take praw.comment as argument with guarantee that there is at least one occurence of !benchmark . It would have to probably regex match with "\\!benchmark (.{0,10}) (i7|i5|i3)-(\\d)(X|K)" or something like that. Can't write more specific without seeing possible data. And also this function would obviously return model names in proper syntax (just like they are written in dictionary)

Now with this set up main routine could look like this:

reddit_client = praw.Reddit(user_agent='<your user agent>'
#OAuth2 shenanigans here
d = {} #filled dictionary
answeredComments = []
while True:
    for comment in reddit_client.get_comments('subreddit'):
        if '!benchmark' not in comment.body.lower():
            continue
        if comment in answeredComments:
            continue
        models = parse_models_from_comment(comment)
        if len(models)==0:
            response = 'I was unable to find benchamrk for given query'
        else:
            response = 'Here are some benchmarks for:\n\n'
            for model in models:
                response += str(model) + ' ' + str(d[model])+'\n\n'
        save_this_comment(comment)
        comment.reply(response)
    time.sleep(900) #15 min break

So on top of stuff that I've written earlier, here is code explanation.

reddit_client.get_comments('subreddit') obviously returns most recent comments, with default limit of 25, and maximum limit of 1000 (api limitation).

First if checks if this comment is even relevant to bot, if there is not a single !benchmark it will just skip over it. Second check is to make sure this comment has not yet been answered. For purpose of this snippet it's only a local list, but having cache file with answered comments or even single table database (with sqlite or sth easy) will prove necessary. This is used so that bot remembers which comments it already answered. Otherwise it will spam endlessly.

Now we have a comment and a guarantee that bot should be able to find some models. So we run our parse_models_from_comment(comment) function to get list of cpu models that user wishes to get benchmarks for.

If len(models)==0 it means that either user specified cpu that is not in csv, made a typo, inserted wrong format, regex didn't catch it or w/e else. Anyway we don't have any models to work on so we should prepare a reply that bot failed with search.

Otherwise, we have something to work with, and thanks to pre-prepared dictionary for lookups can quickly prepare response with any number of links (could also use markup syntax for that)

When response is prepared, first we need to log to file/database that this comment has been answered and next time praw sends it our way we should skip it.

Lastly we post reply to given comment with prepared response (whether positive or negative). Remember to first log comment to database, then post reply and then optionally check if api request to reply was successful. Other way around might prove buggy.