Gnuplot技术问题：从字符串提取日期配合timecolumn及stats使用

问题背景 #

数据中的日期编码在字符串字段（格式示例：subject_name/20221006085330_test_something），通过自定义函数get_date(c)提取日期部分时，遇到两个问题：

1. 将get_date("string")传入timecolumn()配合stats命令时，触发unknown type in magnitude()错误；
2. 调整脚本后能运行，但统计结果显示11条记录（实际仅10条有效数据）。

初始测试脚本 #

# Test script:
reset session

$DAT<<EOD
# cross
string    date     
subject001_name/20200203144309_session021_id     20200203
subject002_name/20200522103745_session079_id     20200522
subject003_name/20210521172330_session201_id     20210521
subject004_name/20220527130250_session047_id     20220527
subject005_name/20220701171757_session081_id     20220701
subject006_name/20220706102238_session003_id     20220706
subject007_name/20221006110417_session081_id     20221006
subject008_name/20230309144707_session901_id     20230309
subject009_name/20230907080602_session003_id     20230907
subject010_name/20240919143733_session022_id     20240919 
EOD

get_date(c) = substr( strcol(c), strstrt( strcol(c), "/" ) + 1, strstrt( strcol(c), "/" ) + 8 )

blocks              = "cross"
selected_block(i)   = word( blocks, i )

timeFmt             = "%Y%m%d"

stats $DAT i selected_block(1) u ( timecolumn( 2, timeFmt ) ) 

pause -1 

stats $DAT i selected_block(1) u ( timecolumn( get_date( "string" ), timeFmt ) ) 

调整后的测试脚本 #

# Test script:
reset session

$DAT<<EOD
# cross
string    date     
subject001_name/20200203144309_session021_id     20200203 
subject002_name/20200522103745_session079_id     20200522 
subject003_name/20210521172330_session201_id     20210521 
subject004_name/20220527130250_session047_id     20220527 
subject005_name/20220701171757_session081_id     20220701 
subject006_name/20220706102238_session003_id     20220706 
subject007_name/20221006110417_session081_id     20221006 
subject008_name/20230309144707_session901_id     20230309 
subject009_name/20230907080602_session003_id     20230907 
subject010_name/20240919143733_session022_id     20240919 
EOD

get_date(c) = substr( strcol(c), strstrt( strcol(c), "/" ) + 1, strstrt( strcol(c), "/" ) + 8 )

blocks              = "cross"
selected_block(i)   = word( blocks, i )

timeFmt             = "%Y%m%d"
date_column(col)    = strptime( timeFmt, get_date(col) )

stats $DAT i selected_block(1) u ( date_column("string") ) 

问题分析与解决 #

1. 解决unknown type in magnitude()错误 #

timecolumn()的第一个参数要求是列的索引或列名，而get_date("string")返回的是单条记录的日期字符串，不符合参数类型要求，因此触发类型错误。

正确的做法是先将提取到的日期字符串转换为时间戳（用strptime()），再传入stats命令，这也是你调整脚本时的正确方向。

2. 解决统计结果多一条的问题 #

统计结果多一条是因为表头行被纳入统计。数据块中的string date是列名行，stats默认会处理所有行，包括表头。

解决方法有两种：

• 方法一：在stats命令中添加skip 1参数，直接跳过表头行；
• 方法二：优化get_date函数，对不含/的行返回空值，让strptime()生成NaN，stats会自动忽略NaN值。

最终优化脚本

reset session

$DAT<<EOD
# cross
string    date     
subject001_name/20200203144309_session021_id     20200203 
subject002_name/20200522103745_session079_id     20200522 
subject003_name/20210521172330_session201_id     20210521 
subject004_name/20220527130250_session047_id     20220527 
subject005_name/20220701171757_session081_id     20220701 
subject006_name/20220706102238_session003_id     20220706 
subject007_name/20221006110417_session081_id     20221006 
subject008_name/20230309144707_session901_id     20230309 
subject009_name/20230907080602_session003_id     20230907 
subject010_name/20240919143733_session022_id     20240919 
EOD

# 优化后的get_date函数，处理不含/的行返回空值
get_date(c) = (pos = strstrt(strcol(c), "/"), pos > 0 ? substr(strcol(c), pos+1, pos+8) : "")

blocks              = "cross"
selected_block(i)   = word( blocks, i )

timeFmt             = "%Y%m%d"
date_column(col)    = strptime( timeFmt, get_date(col) )

# 添加skip 1确保跳过表头，双重保险
stats $DAT i selected_block(1) skip 1 u ( date_column("string") ) 

内容的提问来源于stack exchange，提问作者magfan