Currently, if messages are generated using bot users, such messages will
crash slack_backup, since user.list API method returns only regular
users. In this commit there is detected a situation, where we have
'bot_id' in the data, and in case there is no such user in local
database, create it using data from bot.info API call.
In case of user replacement there was a bunch of regular expressions
used, which was an overkill. Substituted them with one generic regexp.
Moved method which is responsible for substitution to the base class.
Last change is a cosmetic one - uploaded files are represented as
absolute path with file:// prefix, so that it is easier to open it with
xdg-open for example.
Till now, if we download certain files (like those attached to the
conversation) and we already have the file with the same name, number in
format '%03d' was added just before extension. That way there could be
possibility, that the very same file will be downloaded and stored
multiple times, like:
file.png
file.001.png
file.002.png
...
This commit prevents that by adding comparison between files we already
have and file which is downloaded from slack. Adding another file with
additional number will only have place when stored file and downloaded
have different content.
If comment is sent by the user, different structure of the data is sent.
First of all, the type of this message is "message", but it contain
dictionary under 'comment' key, which can be confusing, which contain
needed data (like user id). For this kind of messages, in case of lack
of 'user' in main dict, dict['comment']['user'] will be used for getting
user identifier, while dict['text'] remains as a message text.
For some reason, database key was treated differently in configparser object
in python 3.4.2. In Python 3.4.5 everything is fine. Fixed the defaults to
make sure all string options are treated equally.
