Fixed bug related to anomaly detection system

README.md

@@ -19,12 +19,11 @@ Zephyr is designed to be simple, fast and efficient, providing only the
 weather data of a given location, without any additional nonsense.
 
 This service communicates through a JSON API, making it suitable for
-any kind of internet-based project or device. I already use it on a widget
+any kind of internet-based project or device. I already use it as a [standalone web app](https://m.marcocetica.com),
-on my phone, on my terminal, on the tmux's status bar, in a couple of
+as a phone widget and on my terminal.
-smart bedside clocks I've built and as a standalone web app.
 
 ## Weather
-As state before, Zephyr talks via HTTP using the JSON format. Therefore, you
+As stated before, Zephyr talks via HTTP using JSON formatting. Therefore, you
 can query it using any HTTP client of your choice. Below you can find some examples
 using `curl`:
 
@@ -130,7 +129,7 @@ As in the previous example, you can append the `i` query parameter to get result
 in imperial units.
 
 ## Wind
-The `/wind/:city` endpoint provides wind related information(such as speed and direction) for a given city:
+The `/wind/:city` endpoint provides wind related information (such as speed and direction) for a given city:
 
 ```sh
 curl -s 'http://127.0.0.1:3000/wind/bolzano' | jq
@@ -238,7 +237,7 @@ curl -s 'http://127.0.0.1:3000/forecast/tapei?h' | jq
 ```
 
 As in the previous examples, you can append the `i` query parameter to get results
-in imperial units(**tip**: you can mix both parameter using `&`).
+in imperial units (**tip**: you can mix both parameter using `&`).
 
 ## Moon
 
@@ -329,7 +328,7 @@ will be flagged as outliers by the model and will be reported as such:
 The anomaly detection algorithm is based on a modified version of the
 [Z-Score](https://en.wikipedia.org/wiki/Standard_score) algorithm, which uses the
 [Median Absolute Deviation](https://en.wikipedia.org/wiki/Median_absolute_deviation) to measure the variability
-in a given sample of quantitative data. The algorithm can be summarized as follows(let $X$ be the sample):
+in a given sample of quantitative data. The algorithm can be summarized as follows (let $X$ be the sample):
 
 
 $$
@@ -374,9 +373,9 @@ but no less than a week, *should* follow a normal distribution.
 
 > [!IMPORTANT]
 > The anomaly detection algorithm works under the assumption that the weather data
-> is normally distributed(at least roughly), this might not be the case on datasets
+> is normally distributed (at least roughly), this might not be the case on datasets
-> with a very small number of samples(e.g. few days of data) or with a large
+> with a very small number of samples (e.g. few days of data) or with a large
-> number of samples(e.g. multi-seasonal data). 
+> number of samples (e.g. multi-seasonal data). 
 
 The algorithm works quite well when these conditions are met, and even with real world data,
 the results were quite satisfactory. However, if it
@@ -399,12 +398,12 @@ if you are using their free tier.
 ## Configuration
 Zephyr requires the following environment variables to be set:
 
-| Variable             | Meaning                                |
+| Variable             | Meaning                                 |
-|----------------------|----------------------------------------|
+|----------------------|---------------------------------------- |
-| `ZEPHYR_ADDR`        | Listen address                         |
+| `ZEPHYR_ADDR`        | Listen address                          |
-| `ZEPHYR_PORT`        | Listen port                            |
+| `ZEPHYR_PORT`        | Listen port                             |
-| `ZEPHYR_TOKEN`       | OpenWeatherMap API key                 |
+| `ZEPHYR_TOKEN`       | OpenWeatherMap API key                  |
-| `ZEPHYR_CACHE_TTL`   | Cache time-to-live(expressed in hours) |
+| `ZEPHYR_CACHE_TTL`   | Cache time-to-live (expressed in hours) |
 
 Each value must be set _before_ launching the application. If you plan to deploy Zephyr using
 Docker, you can specify these variables in the `compose.yml` file.

go.mod

@@ -1,3 +1,3 @@
 module github.com/ceticamarco/zephyr
 
-go 1.24.4
+go 1.25.3

statistics/primitives.go

@@ -45,14 +45,17 @@ func Median(temperatures []float64) float64 {
 		return 0
 	}
 
-	slices.Sort(temperatures)
+	// Sort the array without mutating the original values
-	length := len(temperatures)
+	sortedTemps := slices.Clone(temperatures)
+	slices.Sort(sortedTemps)
+
+	length := len(sortedTemps)
 	midValue := length / 2
 
 	if length%2 == 0 {
-		return (temperatures[midValue-1] + temperatures[midValue]) / 2
+		return (sortedTemps[midValue-1] + sortedTemps[midValue]) / 2
 	} else {
-		return temperatures[midValue]
+		return sortedTemps[midValue]
 	}
 }
 
@@ -63,10 +66,12 @@ func Mode(temperatures []float64) float64 {
 		return 0
 	}
 
-	slices.Sort(temperatures)
+	// Sort the array without mutating the original values
+	sortedTemps := slices.Clone(temperatures)
+	slices.Sort(sortedTemps)
 
 	frequencies := make(map[float64]int)
-	for _, val := range temperatures {
+	for _, val := range sortedTemps {
 		frequencies[val]++
 	}
 
@@ -124,6 +129,7 @@ func RobustZScore(temperatures []float64) []struct {
 		Idx   int
 		Value float64
 	}
+
 	for idx, val := range temperatures {
 		z := scale * (val - med) / madAbsDev