--since option not respected unless --follow is passed as well.

[BryanButlerGit]

When executing the following command from CLI:

chalice logs --since 1h

all logs stored in AWS for the log group name are displayed instead of only the log messages stored in the last hour. So when passing --since on its own, its not respected.

But --since is respected when passing --follow as well:

chalice logs --since 1h --follow

The above works. The last hours worth of logs are returned and displayed.

At a glance, it looks like the issue is located within chalice.awsclient.py.TypedAWSClient.iter_log_events

start_time is being passed as an arg to the method:

    def iter_log_events(
        self,
        log_group_name: str,
        start_time: Optional[datetime] = None,
        interleaved: bool = True,
    ) -> Iterator[CWLogEvent]:
. . .

but its presence is not being detected, converted via botocore.utils.datetime2timestamp and passed as an argument to the paginator.

As a quick test, something like this seems to work for me:

    def iter_log_events(
        self,
        log_group_name: str,
        start_time: Optional[datetime] = None,
        interleaved: bool = True,
    ) -> Iterator[CWLogEvent]:
        logs = self._client('logs')
        paginator = logs.get_paginator('filter_log_events')

# WORKS FOR ME:
        if start_time is not None:
            start_time = int(datetime2timestamp(start_time) * 1000)
            pages = paginator.paginate(
                logGroupName=log_group_name, 
                interleaved=True, 
                startTime=start_time
            )
        else:
            pages = paginator.paginate(
                logGroupName=log_group_name, interleaved=True
            )
# END OF QUICK FIX

        try:
            yield from self._iter_log_messages(pages)
        except logs.exceptions.ResourceNotFoundException:
            # If the lambda function exists but has not been invoked yet,
            # it's possible that the log group does not exist and we'll get
            # a ResourceNotFoundException.  If this happens we return instead
            # of propagating an exception back to the user.
            pass

Or maybe LogEventGenerator.iter_log_events should use TypedAWSClient.filter_log_events instead of TypedAWSClient.iter_log_events, like FollowLogEventGenerator.iter_log_events seems to?

I'm not sure. I only briefly looked at the codebase, so I may be wrong in the fix. In any case, it would be nice if this could be fixed. :-)