#!/usr/bin/env python
# coding: utf-8

# # Job Time Statistics

# In[1]:


import pandas as pd
import subprocess
import sys
import datetime as dt
from io import StringIO


# In[2]:


job_name = 'umui-rerun'
start_date = '2020-02-01'


# The `sacct` command will let us get info for job names:

# In[3]:


cmd = [
    'sacct', '-P', '-o', 'JobID,Account,State,AllocCPUS,CpuTime,CpuTimeRAW,Elapsed,ElapsedRAW',
    '--name', job_name,
    '-S', start_date,
]


# In[4]:


child = subprocess.run(cmd, capture_output=True, encoding='utf8')
if child.returncode:
    print(child.stderr, file=sys.stderr)
    raise ValueError('child failed with code ' + str(child.returncode))

acct_text = child.stdout
acct = pd.read_csv(StringIO(acct_text), sep='|')
acct


# The jobs ending in `.batch` are duplicates - remove:

# In[5]:


jobs = acct[~acct['JobID'].str.endswith('.batch')]
jobs


# What is the CPU time?

# In[6]:


cpu = dt.timedelta(seconds=int(jobs['CPUTimeRAW'].sum()))
cpu


# How many hours?

# In[7]:


cpu.total_seconds() / 3600


# What is the wall clock time?

# In[8]:


wall = dt.timedelta(seconds=int(jobs['ElapsedRaw'].sum()))
wall


# How many hours?

# In[9]:


wall.total_seconds() / 3600


# In[ ]: