嵌入式的库，可方便直接pip安装，不需要单独外面部署
基于dill的pickle，可以方便的将python对象存储
三个嵌入式数据库读写对比，看起来差别并不大，实际上都是本地读写文件，不做索要
NB增加了数据库定长设置和时序数据存储回放，，NB默认是dill的pickle，可以存储函数。这里只展示定长存储和函数存储

In [2]:

from dill import dumps,loads

from unqlite import UnQLite
udb = UnQLite('utmp.db')

from vedis import Vedis
vdb=Vedis('vtmp.db')

from deva import *
ndb = NB('ntmp.db')

In [3]:

# 数据准备
df = NB('naja')['basic_df']
df.describe()

Out[3]:

	pe	totalAssets	timeToMarket
count	9046.000000	9046.000000	9.046000e+03
mean	95.099828	1422.396747	2.007718e+07
std	388.246852	13936.501083	8.237193e+04
min	0.000000	0.320000	1.990121e+07
25%	12.950000	24.787500	2.000093e+07
50%	32.630000	60.340000	2.010013e+07
75%	69.365000	187.610000	2.015032e+07
max	10708.040000	334718.500000	2.020110e+07

In [4]:

%%timeit
vdb['a']=0
vdb['df']=dumps(df)

168 ms ± 15.5 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [5]:

%%timeit
ndb['a']=0
ndb['df']=df

201 ms ± 23.2 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [6]:

%%timeit
udb['a']=0
udb['df']=dumps(df)

182 ms ± 24.6 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [7]:

%%timeit
a = vdb['df']|loads@P

6.29 ms ± 152 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [8]:

%%timeit
b = udb['df']|loads@P

6.73 ms ± 858 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [9]:

%%timeit
c = ndb['df']

6.39 ms ± 88.8 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [10]:

ldb=DBStream('test',maxsize=10,fname='/tmp/ldb')
for i in range(100):
    ldb[i]=1

ldb|len

Out[10]:

In [11]:

ldb.db.popitem()
ldb|len

Out[11]:

In [12]:

foo = lambda x:x**2
ldb['foo']=foo
ldb['foo'](3)

Out[12]:

Related Posts: