git-fit 0.17.8 → 0.17.9

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 63105e40b78dfc16931efa9ffe2b88e9ac53b576f2ce14857401d818d6375c5a
-  data.tar.gz: 792778796644fd2117a9e57b1cb330b2a337ac02010f4297e170639f64b9fa4b
+  metadata.gz: 49d811e8b7771ebd21053d0d274c9ac26b96f8d453bdc627623e5571689c5ccf
+  data.tar.gz: ececedeedda8e83e8689e9a2f63ba227bed19dc430600a385328586a01728b66
 SHA512:
-  metadata.gz: e6787fb194f869eb5dd48226b617e2d9757cbfb9ee6cfc657018f9e0f868f3018bb66b3a63718ef9d919836298ce0f3e612504bf77fc033a2286dcaaf7c9dd35
-  data.tar.gz: 1ef52c06c6a9ff330e7e22dd0bb89152bcbeeb861b430a04ffa1841e8e0280ff6e859e2b12007090784e0e692be2fa0b6a6488b1eae825ca44130a3de2d87a5b
+  metadata.gz: 29d188b5016c849a7770768de9ccd7c82df9755cbcc93e0f0ad98f9461acd85285dc269d3b42018064e9ed030dc309584c53c72161b7b62605f4d5ae864391f7
+  data.tar.gz: bce6161efcb1322a59def53da6636abe4877917b66c4ae420f110d6b7464774537881e64bcade700ae3c47240c1f13c93a07cc4a41660703298d8eee3d583254

data/lib/git_fit/elevation/backfill.rb

@@ -5,6 +5,8 @@ module GitFit
     class Backfill
       SOURCES = %w[igpsport xoss strava garmin garmin_cn xingzhe].freeze
       BATCH_SIZE = 100
+      # Fallback threshold when detail.json elevation is absent (issue #83: baro 2m / GPS 3m).
+      ELEV_THRESHOLD = 2.0
 
       def initialize(db, force: false)
         @db = db
@@ -129,20 +131,47 @@ module GitFit
         end
 
         raw = JSON.parse(File.read(raw_path))
-        alts = raw.dig('stream', 'altitude') || []
+        alts = (raw.dig('stream', 'altitude') || raw.dig('data', 'altitude') || []).map(&:to_f)
         return nil if alts.size < 2
 
+        # detail.json stores the inner API payload ({workout:...}); legacy wraps it in
+        # {code,data:{workout:...},msg}. Prefer xingzhe's own processed elevation.
+        detail = xingzhe_detail_workout(natural_id)
+        dg = detail && detail['elevation_gain']
+        dl = detail && detail['elevation_loss']
+        dm = detail && detail['max_altitude']
+
+        gain = detail_gain?(dg) ? dg : elevation_gain_from_alts(alts, threshold: ELEV_THRESHOLD)
+        loss = detail_gain?(dl) ? dl : elevation_loss_from_alts(alts, threshold: ELEV_THRESHOLD)
+
+        # max: detail max_altitude is reliable (matches stream max in samples); else raw max.
+        # min: stream is per-track baro-offset, so only drop low spikes, keep the bulk.
+        filtered = self.class.filter_low_outliers(alts)
         {
-          elevation_gain: elevation_gain_from_alts(alts)&.round(1),
-          elevation_loss: elevation_loss_from_alts(alts)&.round(1),
-          elevation_min: alts.min&.round(1),
-          elevation_max: alts.max&.round(1),
+          elevation_gain: gain&.to_f&.round(1),
+          elevation_loss: loss&.to_f&.round(1),
+          elevation_min: (filtered.min || alts.min)&.round(1),
+          elevation_max: (dm.is_a?(Numeric) && dm > 0 ? dm : alts.max)&.round(1),
         }
       rescue StandardError => e
         warn "Elevation backfill [xingzhe][#{natural_id}]: #{e.message}"
         nil
       end
 
+      def xingzhe_detail_workout(natural_id)
+        path = File.join('data', 'raw', 'xingzhe', natural_id, 'detail.json')
+        return nil unless File.exist?(path)
+
+        d = JSON.parse(File.read(path))
+        d['workout'] || d.dig('data', 'workout')
+      rescue StandardError
+        nil
+      end
+
+      def detail_gain?(value)
+        value.is_a?(Numeric) && !value.zero?
+      end
+
       def fit_path_for(source, natural_id)
         case source
         when 'garmin', 'garmin_cn'
@@ -152,19 +181,19 @@ module GitFit
         end
       end
 
-      def elevation_gain_from_alts(alts)
+      def elevation_gain_from_alts(alts, threshold: 0.0)
         return 0.0 if alts.size < 2
         (1...alts.size).sum do |i|
           d = alts[i] - alts[i - 1]
-          d > 0 ? d : 0.0
+          d > threshold ? d : 0.0
         end
       end
 
-      def elevation_loss_from_alts(alts)
+      def elevation_loss_from_alts(alts, threshold: 0.0)
         return 0.0 if alts.size < 2
         (1...alts.size).sum do |i|
           d = alts[i - 1] - alts[i]
-          d > 0 ? d : 0.0
+          d > threshold ? d : 0.0
         end
       end
 
@@ -182,6 +211,32 @@ module GitFit
       def skip_row(_run_id)
         @skipped += 1
       end
+
+      class << self
+        # Drop only implausibly-low altitude spikes (baro dropouts), leaving the
+        # genuinely shifted bulk intact. Lower fence = median - factor*1.4826*MAD.
+        def filter_low_outliers(altitudes, factor = 2.0)
+          return altitudes if altitudes.size < 4
+
+          med = median(altitudes)
+          m = mad(altitudes)
+          return altitudes if m.zero?
+
+          fence = med - factor * 1.4826 * m
+          altitudes.reject { |x| x < fence }
+        end
+
+        def median(values)
+          s = values.sort
+          n = s.size
+          n.odd? ? s[n / 2] : (s[n / 2 - 1] + s[n / 2]) / 2.0
+        end
+
+        def mad(values)
+          med = median(values)
+          median(values.map { |x| (x - med).abs })
+        end
+      end
     end
   end
 end

data/lib/git_fit/sync/xingzhe.rb

@@ -352,8 +352,9 @@ module GitFit
       end
 
       def derive_elevation_min(stream, detail_workout)
-        if stream && (alt = stream['altitude']) && alt.is_a?(Array) && alt.any?
-          v = alt.compact.min
+        if stream && (alt = stream['altitude']) && alt.is_a?(Array) && alt.size >= 2
+          alts = alt.compact.map(&:to_f)
+          v = GitFit::Elevation::Backfill.filter_low_outliers(alts).min
           return v.to_f if v
         end
         if detail_workout && (segs = detail_workout['segments_km']) && segs.is_a?(Array) && segs.any?

data/lib/git_fit/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module GitFit
-  VERSION = '0.17.8'
+  VERSION = '0.17.9'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: git-fit
 version: !ruby/object:Gem::Version
-  version: 0.17.8
+  version: 0.17.9
 platform: ruby
 authors:
 - Lax