mirror of
https://github.com/LC044/WeChatMsg
synced 2024-11-12 20:21:21 +08:00
add Explain to func
This commit is contained in:
parent
c40db420a1
commit
b0c5ef466d
@ -22,6 +22,18 @@ wordcloud_height = 720
|
|||||||
|
|
||||||
|
|
||||||
def wordcloud(wxid, year='all', who='1'):
|
def wordcloud(wxid, year='all', who='1'):
|
||||||
|
'''
|
||||||
|
词云分析
|
||||||
|
parameters:
|
||||||
|
year: 默认分析全部年份,如果自定义,格式为'2023',
|
||||||
|
who: 默认为1,即自己发送的记录,否则为0,对方发送的记录
|
||||||
|
|
||||||
|
returns:
|
||||||
|
chart_data: 词云数据
|
||||||
|
keyword: 关键词
|
||||||
|
max_num: 说的最多的数量
|
||||||
|
dialogs:部分关键词对话
|
||||||
|
'''
|
||||||
import jieba
|
import jieba
|
||||||
txt_messages = msg_db.get_messages_by_type(wxid, MsgType.TEXT, year)
|
txt_messages = msg_db.get_messages_by_type(wxid, MsgType.TEXT, year)
|
||||||
if not txt_messages:
|
if not txt_messages:
|
||||||
@ -76,6 +88,19 @@ def wordcloud(wxid, year='all', who='1'):
|
|||||||
|
|
||||||
|
|
||||||
def calendar_chart(wxid, year='all'):
|
def calendar_chart(wxid, year='all'):
|
||||||
|
'''
|
||||||
|
日历图分析
|
||||||
|
parameters:
|
||||||
|
year: 默认分析全部年份,如果自定义,格式为'2023'
|
||||||
|
|
||||||
|
returns:
|
||||||
|
chart_data: 图表数据
|
||||||
|
data_length: 和对方的聊天记录总数
|
||||||
|
max_date: 聊天最多的一天日期
|
||||||
|
max_num: 聊天最多的一天记录条数
|
||||||
|
date_num: 聊天天数
|
||||||
|
dialogs: 最早聊天记录
|
||||||
|
'''
|
||||||
data_length = msg_db.get_messages_length_with_ta(wxid, year) # 获取和他的聊天条数
|
data_length = msg_db.get_messages_length_with_ta(wxid, year) # 获取和他的聊天条数
|
||||||
print(f'聊天总数:{data_length}')
|
print(f'聊天总数:{data_length}')
|
||||||
calendar_data = msg_db.get_messages_by_days(wxid, year)
|
calendar_data = msg_db.get_messages_by_days(wxid, year)
|
||||||
@ -132,14 +157,25 @@ def calendar_chart(wxid, year='all'):
|
|||||||
'max_date': formatted_date,
|
'max_date': formatted_date,
|
||||||
'max_num': str(max_),
|
'max_num': str(max_),
|
||||||
'date_num': str(date_num),
|
'date_num': str(date_num),
|
||||||
'dialogs': msg_db.get_first_time_of_message(wxid) # 非年度报告使用
|
'dialogs': msg_db.get_first_time_of_message(wxid)
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
||||||
def month_count(wxid, year='all'):
|
def month_count(wxid, year='all'):
|
||||||
"""
|
'''
|
||||||
每月聊天条数
|
每月聊天条数
|
||||||
"""
|
parameters:
|
||||||
|
year: 默认分析全部年份,如果自定义,格式为'2023',
|
||||||
|
|
||||||
|
returns:
|
||||||
|
chart_data: 图表数据
|
||||||
|
txt: 文字描述,
|
||||||
|
month_average_num: 月平均聊天数量
|
||||||
|
max_num_month: 聊天最多的一个月
|
||||||
|
max_num: 聊天最多的月的记录条数
|
||||||
|
min_num_month: 聊天最少的一个月
|
||||||
|
min_num: 聊天最少的月的记录条数
|
||||||
|
'''
|
||||||
msg_data = msg_db.get_messages_by_month(wxid, year)
|
msg_data = msg_db.get_messages_by_month(wxid, year)
|
||||||
y_data = list(map(lambda x: x[1], msg_data))
|
y_data = list(map(lambda x: x[1], msg_data))
|
||||||
x_axis = list(map(lambda x: x[0], msg_data))
|
x_axis = list(map(lambda x: x[0], msg_data))
|
||||||
@ -199,9 +235,18 @@ def month_count(wxid, year='all'):
|
|||||||
|
|
||||||
|
|
||||||
def hour_count(wxid, year='all'):
|
def hour_count(wxid, year='all'):
|
||||||
"""
|
'''
|
||||||
小时计数聊天条数
|
小时计数聊天条数
|
||||||
"""
|
parameters:
|
||||||
|
year: 默认分析全部年份,如果自定义,格式为'2023',
|
||||||
|
|
||||||
|
returns:
|
||||||
|
chart_data: 图表数据
|
||||||
|
max_num_hour: 聊天最多的时间段
|
||||||
|
max_num: 该时间段聊天条数
|
||||||
|
late_data: 聊天晚的时间和聊天数据
|
||||||
|
early_data: 聊天最早的时间和聊天数据
|
||||||
|
'''
|
||||||
msg_data = msg_db.get_messages_by_hour(wxid, year)
|
msg_data = msg_db.get_messages_by_hour(wxid, year)
|
||||||
print(msg_data)
|
print(msg_data)
|
||||||
y_data = list(map(lambda x: x[1], msg_data))
|
y_data = list(map(lambda x: x[1], msg_data))
|
||||||
@ -250,6 +295,19 @@ def hour_count(wxid, year='all'):
|
|||||||
|
|
||||||
|
|
||||||
def emoji_count(wxid, year='all'):
|
def emoji_count(wxid, year='all'):
|
||||||
|
'''
|
||||||
|
表情和表情包统计
|
||||||
|
parameters:
|
||||||
|
year: 默认分析全部年份,如果自定义,格式为'2023',
|
||||||
|
|
||||||
|
returns:
|
||||||
|
ta_total_emoji_num: TA发送的表情数量
|
||||||
|
me_total_emoji_num: 我发送的表情数量
|
||||||
|
ta_max_emoji: 他发送的表情数量最多的10个:格式为 [(表情,数量)]
|
||||||
|
me_max_emoji: 我发送的表情数量最多的10个:格式为 [(表情,数量)]
|
||||||
|
MeImgDict: 我常发的3张表情包图片地址+数量,字典格式,path为key
|
||||||
|
MeImgDict: TA常发的3张表情包图片地址+数量,字典格式,path为key
|
||||||
|
'''
|
||||||
# 最常发的表情
|
# 最常发的表情
|
||||||
txt_messages = msg_db.get_messages_by_type(wxid, MsgType.TEXT, year)
|
txt_messages = msg_db.get_messages_by_type(wxid, MsgType.TEXT, year)
|
||||||
me_txt_messages = ''.join(map(lambda x: x[7] if x[4] == 1 else '', txt_messages))
|
me_txt_messages = ''.join(map(lambda x: x[7] if x[4] == 1 else '', txt_messages))
|
||||||
|
Loading…
Reference in New Issue
Block a user