Merge pull request #2 from akshay-ranganath/main

Adding queries from past year
eeeps · Nov 1, 2021 · 2fe4cfc · 2fe4cfc
2 parents 0ad2835 + 9db8161
commit 2fe4cfc
Show file tree

Hide file tree

Showing 24 changed files with 1,022 additions and 0 deletions.
diff --git a/sql/2021/media/big_non_custom_metrics.sql b/sql/2021/media/big_non_custom_metrics.sql
@@ -0,0 +1,37 @@
+#standardSQL
+# non custom metrics sql that uses regexp on response bodies
+# img src vs data-uri
+# count rel=preconnect
+# video with src
+# video with source
+# figure
+# figure with figcaption
+
+SELECT
+  client,
+  COUNTIF(has_img_data_uri) / COUNTIF(has_img_src) AS pages_with_img_data_uri_pct,
+  COUNTIF(rel_preconnect) / COUNT(0) AS pages_with_rel_preconnect_pct,
+  COUNTIF(has_video_src) / COUNT(0) AS pages_with_video_src_pct,
+  COUNTIF(has_video_source) / COUNT(0) AS pages_with_video_source_pct,
+  COUNTIF(has_figure) / COUNT(0) AS pages_with_figure_pct,
+  COUNTIF(has_figcaption) / COUNT(0) AS pages_with_figcaption_pct
+FROM (
+  SELECT
+    client,
+    page,
+    REGEXP_CONTAINS(body, r'(?i)<img[^><]*src=(?:\"|\')*data[:]image/(?:\"|\')*[^><]*>') AS has_img_data_uri,
+    REGEXP_CONTAINS(body, r'(?i)<img[^><]*src=[^><]*>') AS has_img_src,
+    REGEXP_CONTAINS(body, r'(?i)<link[^><]*rel=(?:\"|\')*preconnect/(?:\"|\')*[^><]*>') AS rel_preconnect,
+    REGEXP_CONTAINS(body, r'(?i)<video[^><]*src=[^><]*>') AS has_video_src,
+    REGEXP_CONTAINS(body, r'(?i)<video[^><]*>.*?<source[^><]*>.*?</video>') AS has_video_source,
+    REGEXP_CONTAINS(body, r'(?i)<figure[^><]*>') AS has_figure,
+    REGEXP_CONTAINS(body, r'(?i)<figure[^><]*>.*?<figcaption[^><]*>.*?</figure>') AS has_figcaption
+  FROM
+    `httparchive.almanac.summary_response_bodies`
+  WHERE
+    date = '2021-08-01' AND
+    firstHtml)
+GROUP BY
+  client
+ORDER BY
+  client
diff --git a/sql/2021/media/bytes_per_pixel.sql b/sql/2021/media/bytes_per_pixel.sql
@@ -0,0 +1,89 @@
+#standardSQL
+# Bytes per pixel per image format
+CREATE TEMPORARY FUNCTION getImages(payload STRING)
+RETURNS ARRAY<STRUCT<url STRING, naturalWidth INT64, naturalHeight INT64, width INT64, height INT64>> LANGUAGE js AS '''
+try {
+  var $ = JSON.parse(payload);
+  var images = JSON.parse($._Images) || [];
+  return images.map(({url, naturalHeight, naturalWidth, width, height}) => ({url, naturalHeight: Number.parseInt(naturalHeight) || 0, naturalWidth: Number.parseInt(naturalWidth) || 0, width: Number.parseInt(width) || 0, height: Number.parseInt(height) || 0}));
+} catch (e) {}
+return null;
+''';
+
+SELECT
+  a.client,
+  imageType,
+  count(0) AS count,
+  APPROX_QUANTILES(if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 1000)[OFFSET(100)] AS pixels_p10,
+  APPROX_QUANTILES(if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 1000)[OFFSET(250)] AS pixels_p25,
+  APPROX_QUANTILES(if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 1000)[OFFSET(500)] AS pixels_p50,
+  APPROX_QUANTILES(if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 1000)[OFFSET(750)] AS pixels_p75,
+  APPROX_QUANTILES(if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 1000)[OFFSET(900)] AS pixels_p90,
+  APPROX_QUANTILES(bytes, 1000)[OFFSET(100)] AS bytes_p10,
+  APPROX_QUANTILES(bytes, 1000)[OFFSET(250)] AS bytes_p25,
+  APPROX_QUANTILES(bytes, 1000)[OFFSET(500)] AS bytes_p50,
+  APPROX_QUANTILES(bytes, 1000)[OFFSET(750)] AS bytes_p75,
+  APPROX_QUANTILES(bytes, 1000)[OFFSET(900)] AS bytes_p90,
+  APPROX_QUANTILES(round(bytes/if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 4), 1000)[OFFSET(100)] AS bpp_p10,
+  APPROX_QUANTILES(round(bytes/if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 4), 1000)[OFFSET(250)] AS bpp_p25,
+  APPROX_QUANTILES(round(bytes/if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 4), 1000)[OFFSET(500)] AS bpp_p50,
+  APPROX_QUANTILES(round(bytes/if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 4), 1000)[OFFSET(750)] AS bpp_p75,
+  APPROX_QUANTILES(round(bytes/if(imageType = 'svg' AND pixels > 0, pixels, naturalPixels), 4), 1000)[OFFSET(900)] AS bpp_p90
+FROM
+(
+  SELECT
+    _TABLE_SUFFIX AS client,
+    p.url AS page,
+    image.url AS url,
+    image.width AS width,
+    image.height AS height,
+    image.naturalWidth AS naturalWidth,
+    image.naturalHeight AS naturalHeight,
+    IFNULL(image.width, 0) * IFNULL(image.height, 0) AS pixels,
+    IFNULL(image.naturalWidth, 0) * IFNULL(image.naturalHeight, 0) AS naturalPixels
+  FROM
+    `httparchive.pages.2021_08_01_*` p
+    CROSS JOIN UNNEST(getImages(payload)) AS image
+  WHERE
+    image.naturalHeight > 0 AND
+    image.naturalWidth > 0
+) a
+LEFT JOIN
+(
+    SELECT
+      client,
+      page,
+      url,
+      NULLIF(if(REGEXP_CONTAINS(mimetype, r'(?i)^application|^applicaton|^binary|^image$|^multipart|^media|^$|^text/html|^text/plain|\d|array|unknown|undefined|\*|string|^img|^images|^text|\%2f|\(|ipg$|jpe$|jfif'), format, LOWER(regexp_replace(regexp_replace(mimetype, r'(?is).*image[/\\](?:x-)?|[\."]|[ +,;]+.*$', ''), r'(?i)pjpeg|jpeg', 'jpg'))), '') AS imageType,
+      respSize AS bytes
+    FROM `httparchive.almanac.requests`
+
+    WHERE
+      # many 404s and redirects show up as image/gif
+      status = 200 AND
+
+      # we are trying to catch images. WPO populates the format for media but it uses a file extension guess.
+      #So we exclude mimetypes that aren't image or where the format couldn't be guessed by WPO
+     (format <> '' OR mimetype LIKE 'image%') AND
+
+      # many image/gifs are really beacons with 1x1 pixel, but svgs can get caught in the mix
+     (respSize > 1500 OR REGEXP_CONTAINS(mimetype, r'svg')) AND
+
+      # strip favicon requests
+     format <> 'ico' AND
+
+      # strip video mimetypes and other favicons
+      NOT REGEXP_CONTAINS(mimetype, r'video|ico')
+) b
+ON (b.client = a.client AND a.page = b.page AND a.url = b.url)
+
+WHERE
+  naturalPixels > 0 AND
+  bytes > 0 AND
+  imageType IN ('jpg', 'png', 'webp', 'gif', 'svg')
+GROUP BY
+  client,
+  imageType
+ORDER BY
+  client DESC
+  imageType DESC;
diff --git a/sql/2021/media/image_alt.sql b/sql/2021/media/image_alt.sql
@@ -0,0 +1,52 @@
+#standardSQL
+# usage of alt text in images
+
+# returns all the data we need from _markup
+CREATE TEMPORARY FUNCTION get_markup_info(markup_string STRING)
+RETURNS STRUCT<
+  total INT64,
+  alt_missing INT64, 
+  alt_blank INT64,
+  alt_present INT64,
+  decode_lazy INT64
+
+> LANGUAGE js AS '''
+var result = {};
+try {
+    var markup = JSON.parse(markup_string);
+
+    if (Array.isArray(markup) || typeof markup != 'object') return result;
+	
+    result.total = markup.images.img.total;
+    result.alt_missing = markup.images.img.alt.missing;
+    result.alt_blank = markup.images.img.alt.blank;
+    result.alt_present = markup.images.img.alt.present;
+    result.decode_lazy = markup.images.img.decoding || 0;
+
+} catch (e) {}
+return result;
+''';
+
+SELECT
+  client,
+  SAFE_DIVIDE(COUNTIF(markup_info.total > 0), COUNT(0)) AS pages_with_img_pct,
+  SAFE_DIVIDE(COUNTIF(markup_info.alt_missing > 0), COUNT(0)) AS pages_with_alt_missing_pct,
+  SAFE_DIVIDE(COUNTIF(markup_info.alt_blank > 0), COUNT(0)) AS pages_with_alt_blank_pct,
+  SAFE_DIVIDE(COUNTIF(markup_info.alt_present > 0), COUNT(0)) AS pages_with_alt_present_pct,
+  SAFE_DIVIDE(COUNTIF(markup_info.decode_lazy > 0), COUNT(0)) as pages_with_decode_pct,
+  SUM(markup_info.total) AS img_total,
+  SAFE_DIVIDE(SUM(markup_info.alt_missing), SUM(markup_info.total)) AS imgs_alt_missing_pct,
+  SAFE_DIVIDE(SUM(markup_info.alt_blank), SUM(markup_info.total)) AS img_alt_blank_pct,
+  SAFE_DIVIDE(SUM(markup_info.alt_present), SUM(markup_info.total)) AS img_alt_present_pct,
+  SAFE_DIVIDE(SUM(markup_info.decode_lazy), SUM(markup_info.total)) AS img_decode_pct
+FROM (
+  SELECT
+    _TABLE_SUFFIX AS client,
+    url,
+    get_markup_info(JSON_EXTRACT_SCALAR(payload, '$._markup')) AS markup_info
+  FROM
+    `httparchive.pages.2021_08_01_*`)
+GROUP BY
+  client
+ORDER BY
+  client
diff --git a/sql/2021/media/image_dimension_by_device.sql b/sql/2021/media/image_dimension_by_device.sql
@@ -0,0 +1,37 @@
+#standardSQL
+# Taken from Performance 2020 chapter
+# Removed the condition to look for only ECommerce
+
+## $4.11 run
+
+CREATE TEMPORARY FUNCTION getImageDimensions(payload STRING)
+RETURNS ARRAY<STRUCT<height INT64, width INT64>> LANGUAGE js AS '''
+try {
+  var $ = JSON.parse(payload);
+  var images = JSON.parse($._Images);
+  return images.map(i => ({height: i.naturalHeight, width: i.naturalWidth}));
+} catch (e) {
+  return [];
+}
+''';
+
+SELECT
+  percentile,
+  category,
+  _TABLE_SUFFIX AS client,
+  APPROX_QUANTILES(image.width, 1000)[OFFSET(percentile * 10)] AS image_width,
+  APPROX_QUANTILES(image.height, 1000)[OFFSET(percentile * 10)] AS image_height
+FROM
+  `httparchive.pages.2021_08_01_*`
+JOIN
+  `httparchive.technologies.2021_08_01_*`
+USING (_TABLE_SUFFIX, url),
+UNNEST(getImageDimensions(payload)) AS image,
+UNNEST([10, 25, 50, 75, 90]) AS percentile
+GROUP BY
+  percentile,
+  client,
+  category
+ORDER BY
+  percentile,
+  client
diff --git a/sql/2021/media/image_mimetype_ext.sql b/sql/2021/media/image_mimetype_ext.sql
@@ -0,0 +1,25 @@
+#standardSQL
+# images mimetype vs extension
+SELECT
+  client,
+  ext,
+  mimetype,
+  COUNT(0) AS ext_mime_image_count,
+  SUM(COUNT(0)) OVER (PARTITION BY client) AS total_images,
+  SAFE_DIVIDE(COUNT(0), SUM(COUNT(0)) OVER (PARTITION BY client)) AS total_image_pct,
+  SAFE_DIVIDE(COUNT(0), SUM(COUNT(0)) OVER (PARTITION BY client, ext)) AS ext_pct,
+  SAFE_DIVIDE(COUNT(0), SUM(COUNT(0)) OVER (PARTITION BY client, mimetype)) AS mime_pct
+FROM
+  `httparchive.almanac.requests`
+WHERE
+  date = '2021-08-01' and
+  type = 'image'
+GROUP BY
+  client,
+  ext,
+  mimetype
+HAVING
+  ext_mime_image_count > 10000
+ORDER BY
+  ext_mime_image_count DESC,
+  client
diff --git a/sql/2021/media/image_srcset_candidates.sql b/sql/2021/media/image_srcset_candidates.sql
@@ -0,0 +1,41 @@
+#standardSQL
+# images srcset candidates average
+
+# returns all the data we need from _media
+CREATE TEMPORARY FUNCTION get_media_info(media_string STRING)
+RETURNS STRUCT<
+  num_srcset_all INT64, 
+  num_srcset_candidates_avg INT64
+> LANGUAGE js AS '''
+var result = {};
+try {
+    var media = JSON.parse(media_string);
+
+    if (Array.isArray(media) || typeof media != 'object') return result;
+	
+    result.num_srcset_all = media.num_srcset_all;
+    result.num_srcset_candidates_avg = 
+	    media.num_srcset_all == 0? 0: (media.num_srcset_candidates / media.num_srcset_all);
+
+} catch (e) {}
+return result;
+''';
+
+SELECT
+  client,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_all > 0), COUNT(0)) AS pages_with_srcset_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_candidates_avg >= 1 AND media_info.num_srcset_candidates_avg <=3), COUNTIF(media_info.num_srcset_all > 0)) AS pages_with_srcset_candidates_1_3_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_candidates_avg >= 1 AND media_info.num_srcset_candidates_avg <= 5), COUNTIF(media_info.num_srcset_all > 0)) AS pages_with_srcset_candidates_1_5_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_candidates_avg > 5 AND media_info.num_srcset_candidates_avg <= 10), COUNTIF(media_info.num_srcset_all > 0)) AS pages_with_srcset_candidates_5_10_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_candidates_avg > 10 AND media_info.num_srcset_candidates_avg <= 15), COUNTIF(media_info.num_srcset_all > 0)) AS pages_with_srcset_candidates_10_15_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_candidates_avg > 15 AND media_info.num_srcset_candidates_avg <= 20), COUNTIF(media_info.num_srcset_all > 0)) AS pages_with_srcset_candidates_15_20_pct
+FROM (
+  SELECT
+    _TABLE_SUFFIX AS client,
+    get_media_info(JSON_EXTRACT_SCALAR(payload, '$._media')) AS media_info
+  FROM
+    `httparchive.pages.2021_08_01_*`)
+GROUP BY
+  client
+ORDER BY
+  client
diff --git a/sql/2021/media/image_srcset_descriptor.sql b/sql/2021/media/image_srcset_descriptor.sql
@@ -0,0 +1,41 @@
+#standardSQL
+# images with srcset descriptor_x descriptor_w
+
+# returns all the data we need from _media
+CREATE TEMPORARY FUNCTION get_media_info(media_string STRING)
+RETURNS STRUCT<
+  num_srcset_all INT64, 
+  num_srcset_descriptor_x INT64,
+  num_srcset_descriptor_w INT64
+> LANGUAGE js AS '''
+var result = {};
+try {
+    var media = JSON.parse(media_string);
+
+    if (Array.isArray(media) || typeof media != 'object') return result;
+	
+    result.num_srcset_all = media.num_srcset_all;
+    result.num_srcset_descriptor_x = media.num_srcset_descriptor_x;
+	result.num_srcset_descriptor_w = media.num_srcset_descriptor_w;
+
+} catch (e) {}
+return result;
+''';
+
+SELECT
+  client,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_all > 0), COUNT(0)) AS pages_with_srcset_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_descriptor_x > 0), COUNT(0)) AS pages_with_srcset_descriptor_x_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_descriptor_w > 0), COUNT(0)) AS pages_with_srcset_descriptor_w_pct,
+  SAFE_DIVIDE(SUM(media_info.num_srcset_descriptor_x), SUM(media_info.num_srcset_all)) AS instances_of_srcset_descriptor_x_pct,
+  SAFE_DIVIDE(SUM(media_info.num_srcset_descriptor_w), SUM(media_info.num_srcset_all)) AS instances_of_srcset_descriptor_w_pct
+FROM (
+  SELECT
+    _TABLE_SUFFIX AS client,
+    get_media_info(JSON_EXTRACT_SCALAR(payload, '$._media')) AS media_info
+  FROM
+    `httparchive.pages.2021_08_01_*`)
+GROUP BY
+  client
+ORDER BY
+  client
diff --git a/sql/2021/media/image_srcset_sizes.sql b/sql/2021/media/image_srcset_sizes.sql
@@ -0,0 +1,39 @@
+#standardSQL
+# images with srcset w/wo sizes
+
+# returns all the data we need from _media
+CREATE TEMPORARY FUNCTION get_media_info(media_string STRING)
+RETURNS STRUCT<
+  num_srcset_all INT64, 
+  num_srcset_sizes INT64
+> LANGUAGE js AS '''
+var result = {};
+try {
+    var media = JSON.parse(media_string);
+
+    if (Array.isArray(media) || typeof media != 'object') return result;
+	
+    result.num_srcset_all = media.num_srcset_all;
+    result.num_srcset_sizes = media.num_srcset_sizes;
+
+} catch (e) {}
+return result;
+''';
+
+SELECT
+  client,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_all > 0), COUNT(0)) AS pages_with_srcset_pct,
+  SAFE_DIVIDE(COUNTIF(media_info.num_srcset_sizes > 0), COUNT(0)) AS pages_with_srcset_sizes_pct,
+  SAFE_DIVIDE((COUNTIF(media_info.num_srcset_all > 0) - COUNTIF(media_info.num_srcset_sizes > 0)), COUNT(0)) AS pages_with_srcset_wo_sizes_pct,
+  SAFE_DIVIDE(SUM(media_info.num_srcset_sizes), SUM(media_info.num_srcset_all)) AS instances_of_srcset_sizes_pct,
+  SAFE_DIVIDE((SUM(media_info.num_srcset_all) - SUM(media_info.num_srcset_sizes)), SUM(media_info.num_srcset_all)) AS instances_of_srcset_wo_sizes_pct
+FROM (
+  SELECT
+    _TABLE_SUFFIX AS client,
+    get_media_info(JSON_EXTRACT_SCALAR(payload, '$._media')) AS media_info
+  FROM
+    `httparchive.pages.2021_08_01_*`)
+GROUP BY
+  client
+ORDER BY
+  client
diff --git a/sql/2021/media/img_covid_in_path.sql b/sql/2021/media/img_covid_in_path.sql
@@ -0,0 +1,22 @@
+#standardSQL
+# images with covid in path - average size
+SELECT
+  client,
+  LOWER(ext) AS ext,
+  COUNT(0) AS ext_count,
+  COUNTIF(REGEXP_CONTAINS(LOWER(url), r'[^/]*?[:]//[^/]*?/.*?covid')) AS ext_count_covid,
+  SAFE_DIVIDE(SUM(respSize), COUNT(0)) AS avg_size,
+  SAFE_DIVIDE(SUM(IF(REGEXP_CONTAINS(LOWER(url), r'[^/]*?[:]//[^/]*?/.*?covid'), respSize, 0)), COUNTIF(REGEXP_CONTAINS(LOWER(url), r'[^/]*?[:]//[^/]*?/.*?covid'))) AS avg_size_covid
+FROM
+  `httparchive.almanac.requests`
+WHERE
+  date = '2021-08-01' AND
+  type = 'image'
+GROUP BY
+  client,
+  ext
+HAVING
+  ext_count_covid > 100
+ORDER BY
+  client,
+  ext_count_covid DESC;