我想根据一组间隔从vector中排除值。
示例数据：
<#
.SYNOPSIS
Test script for the ParameterizedPropertyAccessor object.
#>
<#
.SYNOPSIS
Create a new PSCustomObject which will contain a NoteProperty called Item accessed like a ParameterizedProperty.
#>
Function New-TestPPA
{
# Instantiate our test object.
$testPPA = New-Object -TypeName PSCustomObject;
# Create a new instance of our PPA object, added to our test object, providing it Get and Set script blocks.
# Note that currently the scripts are set up for Powershell 4 and above. If you are using a version of Powershell
# previous to this, comment out the current Param() values, and uncomment the alternate Param() values.
$ppa = New-ParameterizedPropertyAccessor -Parent $testPPA -Name Item -Get `
{
Param(
<#
[Parameter(Mandatory = $true, Position = 0)]
[PSObject] $this,
[Parameter(Mandatory = $true, Position = 1)]
[string] $Key
#>
[Parameter(Mandatory = $true, Position = 0)]
[string] $Key
)
$this._ht[$Key];
} -Set {
Param(
<#
[Parameter(Mandatory = $true, Position = 0)]
[PSObject] $this,
[Parameter(Mandatory = $true, Position = 1)]
[string] $Key,
[Parameter(Mandatory = $true, Position = 2)]
[string] $Value
#>
[Parameter(Mandatory = $true, Position = 0)]
[string] $Key,
[Parameter(Mandatory = $true, Position = 1)]
[string] $Value
)
$this._ht[$Key] = $Value;
};
# Add a HashTable <_ht> used as our backing store. Note that this could be any keyed collection type object.
$testPPA | Add-Member -MemberType NoteProperty -Name _ht -Value @{} -PassThru;
}
[string] $scriptDir = Split-Path -Path $MyInvocation.MyCommand.Definition -Parent;
Import-Module $scriptDir\PSObjectWrappers.psm1;
# Create test object.
$testPPA = New-TestPPA;
# Note that "Item" property is actually a NoteProperty of type ParameterizedPropertyAccessor.
Write-Host "Type '`$testPPA | gm' to see Item NoteProperty.";
# Note that it is the ParameterizedPropertyAccessor object retrieved that has a ParameterizedProperty.
# Also note that Powershell has named this property "Item".
Write-Host "Type '`$testPPA.Item | gm' to see Item ParameterizedProperty";
# Step through what happens when we "set" the "parameterized" Item property.
# Note that this is actually retrieving the Item NoteProperty, and then setting its default accessor, which calls
# the 'Set' ScriptBlock.
Write-Host "";
Write-Host "Setting Name value";
Write-Host "... to 'Mark'."
$testPPA.Item["Name"] = "Mark";
# Step through what happens when we "get" the "parameterized" Item property.
# Note that this is actually retrieving the Item NoteProperty, and then retrieving its default accessor, which calls
# the 'Get' ScriptBlock.
Write-Host "";
Write-Host "Retrieving Name value:";
$temp = $testPPA.Item["Name"];
Write-Host $temp;
解决方案1：使用简单的子集（） - 不合适 - mIntervals的长度可能非常大
解决方案2：使用嵌套for循环：
mydata <- sort(runif(100,0,300))
mIntervals <- data.frame(start = c(2,50,97,159) , end = c(5,75, 120, 160))
此解决方案在R中耗时太长。
解决方案3：函数findIntervals
valid <- vector(length(mydata))
valid <- TRUE
for(i in 1:length(mydata){
for(j in 1:length(mIntervals){
if(mydata[i] > mIntervals[j,]$start & mydata[i] < mIntervals[j,]$end){
valid[i] <- FALSE
}
}
}
mydata[valid]
解决方案4：以某种方式使用包'间隔'，但也没有合适的功能（可能是interval_overlap（））
已经讨论了相似（但不完全相同）的问题here。但是有整数向量的解决方案，而不是连续变量。
我没有更多的想法。解决方案号3似乎是最好的，但我不喜欢它 - 它不健壮 - 你必须检查重叠间隔等。
对于这个非常简单的问题，有没有更好的解决方案？ THX
真实数据：我有时会测量光强度（日期时间，强度）。我还有测量设备维护（开始，结束）的日期时间间隔。 现在我想清理数据=排除在维护期间测量的值（有效！）。
答案 0 :(得分：8)
使用data.table的{{3}}，我们可以尝试%anywhere%：
library(data.table)
# %anywhere% returns TRUE if mydata is within any mIntervals, else FALSE
ans <- mydata[!mydata %anywhere% mIntervals]
这将包括端点，但incbounds = TRUE是默认设置。如果需要排除端点，可以使用以下语法：
mydata[!anywhere(mydata, mIntervals[, 1], mIntervals[, 2], incbounds = FALSE)]
答案 1 :(得分：6)
如果重新安排间隔，可以使用cut功能，然后只取出奇数间隔：
NEWinterval <- c(2,5,50,75,97,120,159,160)
mydata[cut(mydata, NEWinterval,labels = F) %% 2 != 0]
答案 2 :(得分：3)
这是一个Rcpp实现：
library(Rcpp);
set.seed(12L);
mydata <- sort(runif(100L,0,300));
mIntervals <- data.frame(start=c(2,50,97,159),end=c(5,75,120,160));
cppFunction('
LogicalVector inIntervals(DoubleVector v, DoubleVector starts, DoubleVector ends ) {
if (starts.size()!=ends.size())
throw new std::invalid_argument("starts and ends must be same length.");
LogicalVector res(v.size(),false);
for (int i = 0; i < v.size(); ++i) {
double val = v[i];
for (int j = 0; j < starts.size(); ++j)
if (val>starts[j] && val<ends[j]) {
res(i) = true;
break;
}
}
return res;
}
');
mydata[!inIntervals(mydata,mIntervals$start,mIntervals$end)];
## [1] 6.863323 10.168687 13.765236 16.585860 20.808275 28.508376 29.355912
## [8] 30.534403 33.809681 37.152610 42.659676 45.787152 46.319152 47.274177
## [15] 47.877135 49.281417 78.640425 79.475513 80.383078 80.814563 88.273175
## [22] 93.344382 94.136411 94.736104 96.603457 126.327013 130.399146 131.800295
## [29] 131.828798 137.282145 148.542361 151.430386 162.212264 162.541752 165.648249
## [36] 166.758025 167.388096 172.243474 172.603380 176.544549 182.477693 189.979382
## [43] 192.404449 192.499610 199.703949 200.945789 202.035664 208.173427 210.533571
## [50] 212.949140 214.431451 215.524016 224.951507 225.608016 229.180120 230.324658
## [57] 232.415456 236.278594 236.350904 244.164168 244.218976 244.669498 245.332560
## [64] 247.184695 253.110672 253.267796 263.339092 263.352697 264.826916 267.979469
## [71] 282.326263 282.786520 285.996158 291.379637 293.290767 294.260683
答案 3 :(得分：2)
我不知道这会有多高效，但是......
vbetween <- Vectorize(dplyr::between, vectorize.args = c("left", "right"), SIMPLIFY=F)
mydata[!Reduce("|", vbetween(mydata, mIntervals$start, mIntervals$end))]
答案 4 :(得分：0)
我想展示另一种使用rolljoin的data.table包的方法。
首先，您融合并订购间隔数据框：
mIntervals.dt <- data.table(mIntervals)
Intervals.melt <- melt(mIntervals.dt, measure.vars = c("start", "end"))
订购数据并使用滚动加入：
mydata.dt <- data.table(mydata)
setkey(Intervals.melt, value)
setkey(mydata.dt)
final.dt <- Intervals.melt[mydata.dt, roll = -Inf]
仅使用带有“结束”值的数据，因为您使用了-Inf（与mIntervals中的下一个最接近的值合并）。
final.dt[variable == "end"]
非常快速且灵活。