Newer
Older
import os
import re
import subprocess
from pathlib import Path
from typing import List, Literal, Tuple
import polars as pl
import pyarrow.parquet as pq
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
import numpy as np
from .units import as_bytes, convert_si, create_size_bin_labels
from .datetime import *
def parse_scontrol():
job_id = os.getenv('SLURM_JOB_ID')
command = f"scontrol show job {job_id} | grep TRES="
result = subprocess.run(command, shell=True, capture_output=True, text=True).stdout.strip()
tres_pattern=r'.*cpu=(?P<cores>[\d]+),mem=(?P<mem>[\d]+[KMGT]?).*'
cores,mem = re.search(tres_pattern,result).groupdict().values()
cores = int(cores)
mem = convert_si(mem,to_unit='G',use_binary=True)
return [cores,mem]
def as_path(s: str | Path) -> Path:
if not isinstance(s,Path):
s = Path(s)
return s
def prep_size_distribution(
size_bins: int | str | List[int | str] = ['4 kiB','4 MiB','1 GiB','10 GiB','100 GiB','1 TiB'],
**kwargs
) -> Tuple[List[int],List[str]]:
if not isinstance(size_bins,list):
size_bins = [size_bins]
size_bins = [as_bytes(s) if isinstance(s,str) else s for s in size_bins]
size_bins = list(set(size_bins))
size_bins.sort() # Sorts and removes any duplicates
size_bins = [s for s in size_bins if s > 0] # Removes 0, as it will be implicit as the left-most break point
size_labels = create_size_bin_labels(size_bins)
return size_bins,size_labels
def calculate_size_distribution(
sizes: pl.Series,
size_bins: int | str | List[int | str] = ['4 kiB','4 MiB','1 GiB','10 GiB','100 GiB','1 TiB'],
**kwargs
) -> pl.Series:
size_bins,size_labels = prep_size_distribution(size_bins)
size_grps = (
sizes
.cut(
breaks=size_bins,
labels=size_labels,
**kwargs
)
.cast(pl.String)
.cast(pl.Enum(size_labels))
)
return size_grps
def prep_age_distribution(
acq: str | np.datetime64,
age_breakpoints: int | List[int],
time_unit: Literal['D','W']
) -> Tuple[List[np.datetime64],List[str]]:
if not isinstance(age_breakpoints,list):
age_breakpoints = [age_breakpoints]
else:
age_breakpoints = list(set(age_breakpoints))
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
age_breakpoints.sort()
age_breakpoints = [t for t in age_breakpoints if t > 0]
# Create age bin labels before converting to duration for easier parsing
age_labels = create_timedelta_labels(age_breakpoints,time_unit)
# # Create age bins by subtracting the number of days from the date
age_breakpoints = create_timedelta_breakpoints(as_datetime(acq),age_breakpoints,time_unit)
return age_breakpoints,age_labels
def calculate_age_distribution(
timestamps: pl.Series,
acq: str | np.datetime64,
age_breakpoints: List[ int ] = [30,60,90,180],
time_unit: Literal['D','W'] = 'D',
**kwargs
) -> pl.Series:
age_breakpoints, age_labels = prep_age_distribution(acq, age_breakpoints, time_unit)
age_grps = (
timestamps
.cut(
breaks=age_breakpoints,
labels=age_labels,
**kwargs
)
.cast(pl.String)
.cast(pl.Enum(age_labels))
)
return age_grps
def get_parquet_dataset_size(parquet_path):
tot_size = 0
for p in parquet_path.glob("*.parquet"):
md = pq.read_metadata(p)
for rg in range(0, md.num_row_groups):
tot_size += md.row_group(rg).total_byte_size
return tot_size