Upload archive analysis and data for catalog realign

We gained an additional 630 matches over previous attempt. None of the images with mismatched archive-manifest IDs matched.
Total number of images missing metadata is 30,746.
3,734 of these are in the train_small subset.

.gitattributes

@@ -67,3 +67,4 @@ eol_realign/data/links_inner.csv filter=lfs diff=lfs merge=lfs -text
 eol_realign/data/links_manifest_cargo_on_md5.csv filter=lfs diff=lfs merge=lfs -text
 eol_realign/data/links_cargo_manifest_IDmismatch.csv filter=lfs diff=lfs merge=lfs -text
 eol_realign/data/eol-cargo-archive_combined-manifest-checksums_links.csv filter=lfs diff=lfs merge=lfs -text
+eol_realign/data/eol-cargo-archive_catalog_combined-manifest-checksums_links.csv filter=lfs diff=lfs merge=lfs -text

eol_realign/data/eol-cargo-archive_catalog_combined-manifest-checksums_links.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc0cc8d4555ed4cf3e5bba2f0c88339bc605ca30982e427d7288e1eb0e74071e
+size 3921517771

eol_realign/notebooks/links_align_reduced.ipynb

@@ -2,18 +2,45 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 1,
    "metadata": {},
    "outputs": [],
    "source": [
     "import pandas as pd"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "TOL_FILEPATH = \"../../data/\""
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "### Read in Links CSV files"
+    "### Read in Links CSV files\n",
+    "\n",
+    "Full cargo archives have some more content than the [original `links_inner` file](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_inner.csv) and the [original `links_manifest_cargo_on_md5.csv`](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_manifest_cargo_on_md5.csv).\n",
+    "\n",
+    "From Matt:\n",
+    "> It represents a series of inner merges:\n",
+    "> 1. The list of images in the EOL cargo archive (used to create the webdatasets) with the EOL portion of `catalog.csv`\n",
+    ">  - Merged on a combined identifier of `eol_content_id` and `eol_page_id`.\n",
+    "> Result from 1) with the list of images redownloaded from all 3 media manifests we touched around the original download time (identified [here](https://huggingface.co/datasets/imageomics/ToL-EDA/discussions/5#6584808d92eba5ad69b8379f)).\n",
+    ">  - Merged on MD5\n",
+    ">\n",
+    "> The result has 6,219,674 unique `treeoflife_id` entries and 6,146,917 unique `md5` entries. It has the combined `eol_content_id` and `eol_page_id` identifiers used at each stage:\n",
+    "- `combined_id_catalog`: represents what's in the original catalog \n",
+    "- `combined_id_manifest_full_manifest`: represents what's in the combined manifest from all 3 dates \n",
+    "> It also has an indicator of which of the three media manifest files each entry was present in (one or more) in columns representing retreival date (eg., `source_0706`: indicates if it was present in the manifest retrieved on July 6, 2023). It also has the copyright owner and license information as contained in the media manifest files.\n",
+    ">\n",
+    "> Still needs to be run through the owner name finding code. ([owner_match](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/71ad5b49226a3a49634931005eeb340b3c96db13/scripts/match_owners.py) --note: this version uses inner merges, so will remove entries without media info. Probably should adjust this.)\n",
+    "\n",
+    "For reference on the numbers, there are 6,250,420 EOL images in TreeOfLife-10M (the catalog). Comparing with `links_inner` previously (from [`links_inner.csv`](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_inner.csv)) we were missing 31,376 images, 29 of which were identified in [`links_manifest_cargo_on_md5.csv`](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_manifest_cargo_on_md5.csv)."
    ]
   },
   {
@@ -22,8 +49,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "links_inner = pd.read_csv(\"../data/links_inner.csv\", low_memory=False)\n",
-    "links_manifest_cargo = pd.read_csv(\"../data/links_manifest_cargo_on_md5.csv\", low_memory=False)"
+    "links_inner = pd.read_csv(\"../data/eol-cargo-archive_catalog_combined-manifest-checksums_links.csv\", low_memory=False)\n",
+    "links_manifest_cargo = pd.read_csv(\"../data/eol-cargo-archive_combined-manifest-checksums_links.csv\", low_memory=False)"
    ]
   },
   {
@@ -36,46 +63,42 @@
      "output_type": "stream",
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
-      "RangeIndex: 6840530 entries, 0 to 6840529\n",
-      "Data columns (total 34 columns):\n",
-      " #   Column                          Non-Null Count    Dtype  \n",
-      "---  ------                          --------------    -----  \n",
-      " 0   eol_content_id                  6840530 non-null  int64  \n",
-      " 1   eol_page_id                     6840530 non-null  int64  \n",
-      " 2   medium_source_url               6840530 non-null  object \n",
-      " 3   eol_full_size_copy_url          6840530 non-null  object \n",
-      " 4   license_name                    6840530 non-null  object \n",
-      " 5   copyright_owner                 6207911 non-null  object \n",
-      " 6   expected_image_filename         6840530 non-null  object \n",
-      " 7   source_0706                     6840530 non-null  bool   \n",
-      " 8   source_0726                     6840530 non-null  bool   \n",
-      " 9   source_1206                     6840530 non-null  bool   \n",
-      " 10  combined_id_manifest            6840530 non-null  object \n",
-      " 11  md5                             6840530 non-null  object \n",
-      " 12  combined_id_manifest_checksums  6840530 non-null  object \n",
-      " 13  eol_content_id_cargo            6840530 non-null  int64  \n",
-      " 14  eol_page_id_cargo               6840530 non-null  int64  \n",
-      " 15  combined_id_cargo               6840530 non-null  object \n",
-      " 16  split                           6840530 non-null  object \n",
-      " 17  treeoflife_id                   6840530 non-null  object \n",
-      " 18  eol_content_id_catalog          6840530 non-null  int64  \n",
-      " 19  eol_page_id_catalog             6840530 non-null  int64  \n",
-      " 20  bioscan_part                    0 non-null        float64\n",
-      " 21  bioscan_filename                0 non-null        float64\n",
-      " 22  inat21_filename                 0 non-null        float64\n",
-      " 23  inat21_cls_name                 0 non-null        float64\n",
-      " 24  inat21_cls_num                  0 non-null        float64\n",
-      " 25  kingdom                         6550801 non-null  object \n",
-      " 26  phylum                          6552534 non-null  object \n",
-      " 27  class                           6531337 non-null  object \n",
-      " 28  order                           6524949 non-null  object \n",
-      " 29  family                          6505404 non-null  object \n",
-      " 30  genus                           6485458 non-null  object \n",
-      " 31  species                         6405635 non-null  object \n",
-      " 32  common                          6840530 non-null  object \n",
-      " 33  combined_id_catalog             6840530 non-null  object \n",
-      "dtypes: bool(3), float64(5), int64(6), object(20)\n",
-      "memory usage: 1.6+ GB\n"
+      "RangeIndex: 6841202 entries, 0 to 6841201\n",
+      "Data columns (total 30 columns):\n",
+      " #   Column                           Non-Null Count    Dtype  \n",
+      "---  ------                           --------------    -----  \n",
+      " 0   filename_archive                 6841202 non-null  object \n",
+      " 1   md5                              6841202 non-null  object \n",
+      " 2   combined_id_archive              6841202 non-null  object \n",
+      " 3   split                            6841202 non-null  object \n",
+      " 4   treeoflife_id                    6841202 non-null  object \n",
+      " 5   bioscan_part                     0 non-null        float64\n",
+      " 6   bioscan_filename                 0 non-null        float64\n",
+      " 7   inat21_filename                  0 non-null        float64\n",
+      " 8   inat21_cls_name                  0 non-null        float64\n",
+      " 9   inat21_cls_num                   0 non-null        float64\n",
+      " 10  kingdom                          6551473 non-null  object \n",
+      " 11  phylum                           6553206 non-null  object \n",
+      " 12  class                            6532009 non-null  object \n",
+      " 13  order                            6525621 non-null  object \n",
+      " 14  family                           6506076 non-null  object \n",
+      " 15  genus                            6486130 non-null  object \n",
+      " 16  species                          6406307 non-null  object \n",
+      " 17  common                           6841202 non-null  object \n",
+      " 18  combined_id_catalog              6841202 non-null  object \n",
+      " 19  filename_manifest                6841202 non-null  object \n",
+      " 20  combined_id_manifest_redownload  6841202 non-null  object \n",
+      " 21  medium_source_url                6841202 non-null  object \n",
+      " 22  eol_full_size_copy_url           6841202 non-null  object \n",
+      " 23  license_name                     6841202 non-null  object \n",
+      " 24  copyright_owner                  6208498 non-null  object \n",
+      " 25  expected_image_filename          6841202 non-null  object \n",
+      " 26  source_0706                      6841202 non-null  bool   \n",
+      " 27  source_0726                      6841202 non-null  bool   \n",
+      " 28  source_1206                      6841202 non-null  bool   \n",
+      " 29  combined_id_full_manifest        6841202 non-null  object \n",
+      "dtypes: bool(3), float64(5), object(22)\n",
+      "memory usage: 1.4+ GB\n"
      ]
     }
    ],
@@ -93,28 +116,29 @@
      "output_type": "stream",
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
-      "RangeIndex: 7513329 entries, 0 to 7513328\n",
-      "Data columns (total 16 columns):\n",
-      " #   Column                          Non-Null Count    Dtype \n",
-      "---  ------                          --------------    ----- \n",
-      " 0   eol_content_id                  7513329 non-null  int64 \n",
-      " 1   eol_page_id                     7513329 non-null  int64 \n",
-      " 2   medium_source_url               7513329 non-null  object\n",
-      " 3   eol_full_size_copy_url          7513329 non-null  object\n",
-      " 4   license_name                    7513329 non-null  object\n",
-      " 5   copyright_owner                 6860238 non-null  object\n",
-      " 6   expected_image_filename         7513329 non-null  object\n",
-      " 7   source_0706                     7513329 non-null  bool  \n",
-      " 8   source_0726                     7513329 non-null  bool  \n",
-      " 9   source_1206                     7513329 non-null  bool  \n",
-      " 10  combined_id_manifest            7513329 non-null  object\n",
-      " 11  md5                             7513329 non-null  object\n",
-      " 12  combined_id_manifest_checksums  7513329 non-null  object\n",
-      " 13  eol_content_id_cargo            7513329 non-null  int64 \n",
-      " 14  eol_page_id_cargo               7513329 non-null  int64 \n",
-      " 15  combined_id_cargo               7513329 non-null  object\n",
-      "dtypes: bool(3), int64(4), object(9)\n",
-      "memory usage: 766.7+ MB\n"
+      "RangeIndex: 7514001 entries, 0 to 7514000\n",
+      "Data columns (total 17 columns):\n",
+      " #   Column                           Non-Null Count    Dtype \n",
+      "---  ------                           --------------    ----- \n",
+      " 0   filename_archive                 7514001 non-null  object\n",
+      " 1   md5_archive                      7514001 non-null  object\n",
+      " 2   combined_id_archive              7514001 non-null  object\n",
+      " 3   filename_manifest                7514001 non-null  object\n",
+      " 4   md5_combined_manifest            7514001 non-null  object\n",
+      " 5   combined_id_manifest_redownload  7514001 non-null  object\n",
+      " 6   eol_content_id                   7514001 non-null  int64 \n",
+      " 7   eol_page_id                      7514001 non-null  int64 \n",
+      " 8   medium_source_url                7514001 non-null  object\n",
+      " 9   eol_full_size_copy_url           7514001 non-null  object\n",
+      " 10  license_name                     7514001 non-null  object\n",
+      " 11  copyright_owner                  6860825 non-null  object\n",
+      " 12  expected_image_filename          7514001 non-null  object\n",
+      " 13  source_0706                      7514001 non-null  bool  \n",
+      " 14  source_0726                      7514001 non-null  bool  \n",
+      " 15  source_1206                      7514001 non-null  bool  \n",
+      " 16  combined_id_full_manifest        7514001 non-null  object\n",
+      "dtypes: bool(3), int64(2), object(12)\n",
+      "memory usage: 824.1+ MB\n"
      ]
     }
    ],
@@ -122,90 +146,174 @@
     "links_manifest_cargo.info(show_counts=True)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "These both have about 600-700 more entries than the earlier files."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 6,
    "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>filename_archive</th>\n",
+       "      <th>md5_archive</th>\n",
+       "      <th>combined_id_archive</th>\n",
+       "      <th>filename_manifest</th>\n",
+       "      <th>md5_combined_manifest</th>\n",
+       "      <th>combined_id_manifest_redownload</th>\n",
+       "      <th>eol_content_id</th>\n",
+       "      <th>eol_page_id</th>\n",
+       "      <th>medium_source_url</th>\n",
+       "      <th>eol_full_size_copy_url</th>\n",
+       "      <th>license_name</th>\n",
+       "      <th>copyright_owner</th>\n",
+       "      <th>expected_image_filename</th>\n",
+       "      <th>source_0706</th>\n",
+       "      <th>source_0726</th>\n",
+       "      <th>source_1206</th>\n",
+       "      <th>combined_id_full_manifest</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "Empty DataFrame\n",
+       "Columns: [filename_archive, md5_archive, combined_id_archive, filename_manifest, md5_combined_manifest, combined_id_manifest_redownload, eol_content_id, eol_page_id, medium_source_url, eol_full_size_copy_url, license_name, copyright_owner, expected_image_filename, source_0706, source_0726, source_1206, combined_id_full_manifest]\n",
+       "Index: []"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# These should be the same\n",
+    "links_manifest_cargo.loc[links_manifest_cargo[\"combined_id_manifest_redownload\"] != links_manifest_cargo[\"combined_id_full_manifest\"]]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
-      "Index: 822516 entries, 93 to 7513188\n",
-      "Data columns (total 16 columns):\n",
-      " #   Column                          Non-Null Count   Dtype \n",
-      "---  ------                          --------------   ----- \n",
-      " 0   eol_content_id                  822516 non-null  int64 \n",
-      " 1   eol_page_id                     822516 non-null  int64 \n",
-      " 2   medium_source_url               822516 non-null  object\n",
-      " 3   eol_full_size_copy_url          822516 non-null  object\n",
-      " 4   license_name                    822516 non-null  object\n",
-      " 5   copyright_owner                 797108 non-null  object\n",
-      " 6   expected_image_filename         822516 non-null  object\n",
-      " 7   source_0706                     822516 non-null  bool  \n",
-      " 8   source_0726                     822516 non-null  bool  \n",
-      " 9   source_1206                     822516 non-null  bool  \n",
-      " 10  combined_id_manifest            822516 non-null  object\n",
-      " 11  md5                             822516 non-null  object\n",
-      " 12  combined_id_manifest_checksums  822516 non-null  object\n",
-      " 13  eol_content_id_cargo            822516 non-null  int64 \n",
-      " 14  eol_page_id_cargo               822516 non-null  int64 \n",
-      " 15  combined_id_cargo               822516 non-null  object\n",
-      "dtypes: bool(3), int64(4), object(9)\n",
-      "memory usage: 90.2+ MB\n"
+      "Index: 822558 entries, 2 to 7513999\n",
+      "Data columns (total 17 columns):\n",
+      " #   Column                           Non-Null Count   Dtype \n",
+      "---  ------                           --------------   ----- \n",
+      " 0   filename_archive                 822558 non-null  object\n",
+      " 1   md5_archive                      822558 non-null  object\n",
+      " 2   combined_id_archive              822558 non-null  object\n",
+      " 3   filename_manifest                822558 non-null  object\n",
+      " 4   md5_combined_manifest            822558 non-null  object\n",
+      " 5   combined_id_manifest_redownload  822558 non-null  object\n",
+      " 6   eol_content_id                   822558 non-null  int64 \n",
+      " 7   eol_page_id                      822558 non-null  int64 \n",
+      " 8   medium_source_url                822558 non-null  object\n",
+      " 9   eol_full_size_copy_url           822558 non-null  object\n",
+      " 10  license_name                     822558 non-null  object\n",
+      " 11  copyright_owner                  797147 non-null  object\n",
+      " 12  expected_image_filename          822558 non-null  object\n",
+      " 13  source_0706                      822558 non-null  bool  \n",
+      " 14  source_0726                      822558 non-null  bool  \n",
+      " 15  source_1206                      822558 non-null  bool  \n",
+      " 16  combined_id_full_manifest        822558 non-null  object\n",
+      "dtypes: bool(3), int64(2), object(12)\n",
+      "memory usage: 96.5+ MB\n"
      ]
     }
    ],
    "source": [
-    "links_mismatch = links_manifest_cargo.loc[links_manifest_cargo[\"combined_id_cargo\"] != links_manifest_cargo[\"combined_id_manifest_checksums\"]]\n",
+    "# archive has replaced cargo\n",
+    "# full manifest or manifest redownload to replace manifest checksums -- they are equal\n",
+    "links_mismatch = links_manifest_cargo.loc[links_manifest_cargo[\"combined_id_archive\"] != links_manifest_cargo[\"combined_id_full_manifest\"]]\n",
     "links_mismatch.info(show_counts = True)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "There are 42 more mismatched entries. Only 13 additional unique images (MD5s)."
+   ]
+  },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 9,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "md5                               624561\n",
-       "combined_id_manifest_checksums    712504\n",
-       "combined_id_cargo                 703409\n",
+       "md5_archive                  624574\n",
+       "combined_id_full_manifest    712540\n",
+       "combined_id_archive          703446\n",
        "dtype: int64"
       ]
      },
-     "execution_count": 7,
+     "execution_count": 9,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
-    "links_mismatch[[\"md5\", \"combined_id_manifest_checksums\", \"combined_id_cargo\"]].nunique()"
+    "links_mismatch[[\"md5_archive\", \"combined_id_full_manifest\", \"combined_id_archive\"]].nunique()"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 10,
    "metadata": {},
    "outputs": [],
    "source": [
-    "links_mismatch.to_csv(\"../data/links_cargo_manifest_IDmismatch.csv\", index = False)"
+    "# This is archive manifest mismatch\n",
+    "links_mismatch.to_csv(\"../data/links_archive_manifest_IDmismatch.csv\", index = False)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 46,
+   "execution_count": 11,
    "metadata": {},
    "outputs": [],
    "source": [
-    "catalog = pd.read_csv(\"../../data/catalog.csv\", low_memory=False)"
+    "catalog = pd.read_csv(TOL_FILEPATH + \"catalog.csv\", low_memory=False)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 47,
+   "execution_count": 12,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -216,7 +324,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 48,
+   "execution_count": 13,
    "metadata": {},
    "outputs": [
     {
@@ -229,7 +337,7 @@
        "dtype: int64"
       ]
      },
-     "execution_count": 48,
+     "execution_count": 13,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -240,7 +348,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 49,
+   "execution_count": 14,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -249,7 +357,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 50,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [
     {
@@ -292,7 +400,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 51,
+   "execution_count": 16,
    "metadata": {},
    "outputs": [
     {
@@ -320,7 +428,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 52,
+   "execution_count": 17,
    "metadata": {
     "lines_to_next_cell": 2
    },
@@ -333,16 +441,16 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 53,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "6219044"
+       "6219674"
       ]
      },
-     "execution_count": 53,
+     "execution_count": 18,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -352,22 +460,29 @@
     "len(matched_catalog_ids)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Here we go, we got an additional 630 matched IDs!"
+   ]
+  },
   {
    "cell_type": "code",
-   "execution_count": 54,
+   "execution_count": 19,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "['100_45516037',\n",
-       " '10000_45511252',\n",
-       " '10001_45509451',\n",
-       " '10002_45509648',\n",
-       " '10003_45510902']"
+       "['29538374_65414274',\n",
+       " '27793900_888015',\n",
+       " '29121641_5618956',\n",
+       " '27596176_607817',\n",
+       " '20300703_267922']"
       ]
      },
-     "execution_count": 54,
+     "execution_count": 19,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -380,26 +495,26 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "So we have about 30K unmatched, let's find those in the catalog and then we'll try to merge with our mismatched cargo & manifest."
+    "We still have about 30K unmatched (as shown below, it's down to 30,746 from 31,376), let's find those in the catalog and then we'll try to merge with our mismatched cargo & manifest."
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 55,
+   "execution_count": 20,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "10017874    20995677_64638554\n",
-       "8964610       21191962_130898\n",
-       "8441264      14142414_1117599\n",
-       "9354726      28866899_2762658\n",
-       "10681937     20538304_2752336\n",
+       "5842839       28875844_579175\n",
+       "7715090       21712402_702458\n",
+       "10112940     27938371_3091824\n",
+       "8606944     27597441_49820955\n",
+       "1247592      14852900_2896866\n",
        "Name: combined_id_catalog, dtype: object"
       ]
      },
-     "execution_count": 55,
+     "execution_count": 20,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -410,7 +525,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 56,
+   "execution_count": 21,
    "metadata": {},
    "outputs": [
     {
@@ -418,25 +533,25 @@
      "output_type": "stream",
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
-      "Index: 31376 entries, 956874 to 10996978\n",
+      "Index: 30746 entries, 956874 to 10996978\n",
       "Data columns (total 13 columns):\n",
       " #   Column                  Non-Null Count  Dtype \n",
       "---  ------                  --------------  ----- \n",
-      " 0   split                   31376 non-null  object\n",
-      " 1   treeoflife_id           31376 non-null  object\n",
-      " 2   eol_content_id_catalog  31376 non-null  int64 \n",
-      " 3   eol_page_id_catalog     31376 non-null  int64 \n",
-      " 4   kingdom                 31203 non-null  object\n",
-      " 5   phylum                  31208 non-null  object\n",
-      " 6   class                   30783 non-null  object\n",
-      " 7   order                   31158 non-null  object\n",
-      " 8   family                  30924 non-null  object\n",
-      " 9   genus                   30822 non-null  object\n",
-      " 10  species                 30029 non-null  object\n",
-      " 11  common                  31376 non-null  object\n",
-      " 12  combined_id_catalog     31376 non-null  object\n",
+      " 0   split                   30746 non-null  object\n",
+      " 1   treeoflife_id           30746 non-null  object\n",
+      " 2   eol_content_id_catalog  30746 non-null  int64 \n",
+      " 3   eol_page_id_catalog     30746 non-null  int64 \n",
+      " 4   kingdom                 30573 non-null  object\n",
+      " 5   phylum                  30578 non-null  object\n",
+      " 6   class                   30153 non-null  object\n",
+      " 7   order                   30528 non-null  object\n",
+      " 8   family                  30294 non-null  object\n",
+      " 9   genus                   30192 non-null  object\n",
+      " 10  species                 29399 non-null  object\n",
+      " 11  common                  30746 non-null  object\n",
+      " 12  combined_id_catalog     30746 non-null  object\n",
       "dtypes: int64(2), object(11)\n",
-      "memory usage: 3.4+ MB\n"
+      "memory usage: 3.3+ MB\n"
      ]
     }
    ],
@@ -447,29 +562,29 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 57,
+   "execution_count": 22,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
        "split                         2\n",
-       "treeoflife_id             31376\n",
-       "eol_content_id_catalog    31376\n",
-       "eol_page_id_catalog        7531\n",
+       "treeoflife_id             30746\n",
+       "eol_content_id_catalog    30746\n",
+       "eol_page_id_catalog        7042\n",
        "kingdom                       7\n",
        "phylum                       33\n",
-       "class                        88\n",
-       "order                       374\n",
-       "family                     1269\n",
-       "genus                      3827\n",
-       "species                    5550\n",
-       "common                     7372\n",
-       "combined_id_catalog       31376\n",
+       "class                        87\n",
+       "order                       355\n",
+       "family                     1165\n",
+       "genus                      3503\n",
+       "species                    5174\n",
+       "common                     6884\n",
+       "combined_id_catalog       30746\n",
        "dtype: int64"
       ]
      },
-     "execution_count": 57,
+     "execution_count": 22,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -482,14 +597,21 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Merge with Mismatched Cargo-Manifest\n",
+    "Seems one class got resolved (equals more orders and down)."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Merge with Mismatched CargoArchive-Manifest\n",
     "\n",
-    "let's merge with `links_mismatch` to see if the mismatched cargo and manifest combined IDs can be linked up."
+    "let's merge with `links_mismatch` to see if the mismatched archive and manifest combined IDs can be linked up."
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 58,
+   "execution_count": 23,
    "metadata": {},
    "outputs": [
     {
@@ -497,72 +619,81 @@
      "output_type": "stream",
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
-      "RangeIndex: 29 entries, 0 to 28\n",
-      "Data columns (total 29 columns):\n",
-      " #   Column                          Non-Null Count  Dtype \n",
-      "---  ------                          --------------  ----- \n",
-      " 0   split                           29 non-null     object\n",
-      " 1   treeoflife_id                   29 non-null     object\n",
-      " 2   eol_content_id_catalog          29 non-null     int64 \n",
-      " 3   eol_page_id_catalog             29 non-null     int64 \n",
-      " 4   kingdom                         29 non-null     object\n",
-      " 5   phylum                          29 non-null     object\n",
-      " 6   class                           29 non-null     object\n",
-      " 7   order                           29 non-null     object\n",
-      " 8   family                          29 non-null     object\n",
-      " 9   genus                           29 non-null     object\n",
-      " 10  species                         29 non-null     object\n",
-      " 11  common                          29 non-null     object\n",
-      " 12  combined_id_catalog             29 non-null     object\n",
-      " 13  eol_content_id                  29 non-null     int64 \n",
-      " 14  eol_page_id                     29 non-null     int64 \n",
-      " 15  medium_source_url               29 non-null     object\n",
-      " 16  eol_full_size_copy_url          29 non-null     object\n",
-      " 17  license_name                    29 non-null     object\n",
-      " 18  copyright_owner                 28 non-null     object\n",
-      " 19  expected_image_filename         29 non-null     object\n",
-      " 20  source_0706                     29 non-null     bool  \n",
-      " 21  source_0726                     29 non-null     bool  \n",
-      " 22  source_1206                     29 non-null     bool  \n",
-      " 23  combined_id_manifest            29 non-null     object\n",
-      " 24  md5                             29 non-null     object\n",
-      " 25  combined_id_manifest_checksums  29 non-null     object\n",
-      " 26  eol_content_id_cargo            29 non-null     int64 \n",
-      " 27  eol_page_id_cargo               29 non-null     int64 \n",
-      " 28  combined_id_cargo               29 non-null     object\n",
-      "dtypes: bool(3), int64(6), object(20)\n",
-      "memory usage: 6.1+ KB\n"
+      "RangeIndex: 0 entries\n",
+      "Data columns (total 30 columns):\n",
+      " #   Column                           Non-Null Count  Dtype \n",
+      "---  ------                           --------------  ----- \n",
+      " 0   split                            0 non-null      object\n",
+      " 1   treeoflife_id                    0 non-null      object\n",
+      " 2   eol_content_id_catalog           0 non-null      int64 \n",
+      " 3   eol_page_id_catalog              0 non-null      int64 \n",
+      " 4   kingdom                          0 non-null      object\n",
+      " 5   phylum                           0 non-null      object\n",
+      " 6   class                            0 non-null      object\n",
+      " 7   order                            0 non-null      object\n",
+      " 8   family                           0 non-null      object\n",
+      " 9   genus                            0 non-null      object\n",
+      " 10  species                          0 non-null      object\n",
+      " 11  common                           0 non-null      object\n",
+      " 12  combined_id_catalog              0 non-null      object\n",
+      " 13  filename_archive                 0 non-null      object\n",
+      " 14  md5_archive                      0 non-null      object\n",
+      " 15  combined_id_archive              0 non-null      object\n",
+      " 16  filename_manifest                0 non-null      object\n",
+      " 17  md5_combined_manifest            0 non-null      object\n",
+      " 18  combined_id_manifest_redownload  0 non-null      object\n",
+      " 19  eol_content_id                   0 non-null      int64 \n",
+      " 20  eol_page_id                      0 non-null      int64 \n",
+      " 21  medium_source_url                0 non-null      object\n",
+      " 22  eol_full_size_copy_url           0 non-null      object\n",
+      " 23  license_name                     0 non-null      object\n",
+      " 24  copyright_owner                  0 non-null      object\n",
+      " 25  expected_image_filename          0 non-null      object\n",
+      " 26  source_0706                      0 non-null      bool  \n",
+      " 27  source_0726                      0 non-null      bool  \n",
+      " 28  source_1206                      0 non-null      bool  \n",
+      " 29  combined_id_full_manifest        0 non-null      object\n",
+      "dtypes: bool(3), int64(4), object(23)\n",
+      "memory usage: 132.0+ bytes\n"
      ]
     }
    ],
    "source": [
+    "# replace manifest checksums with full manifest as above\n",
     "links_catalog_mismatch = pd.merge(mismatched_catalog,\n",
     "                                  links_mismatch,\n",
     "                                  left_on = \"combined_id_catalog\",\n",
-    "                                  right_on = \"combined_id_manifest_checksums\",\n",
+    "                                  right_on = \"combined_id_full_manifest\",\n",
     "                                  how = \"inner\")\n",
     "links_catalog_mismatch.info(show_counts = True)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Now there's no gain from the mismatch file, but that was only 29 images, so we've captured them from the archive (plus more!)."
+   ]
+  },
   {
    "cell_type": "code",
-   "execution_count": 59,
+   "execution_count": 24,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "9251313       9447041_5660435\n",
-       "10368203      28965413_916574\n",
-       "3357181     28940017_45513086\n",
-       "10907940     10517333_3036072\n",
-       "10208533      9444575_1178182\n",
-       "5434229     10517050_46480886\n",
-       "2638483      9446016_46468616\n",
+       "2003559      15270321_240065\n",
+       "8751466       29485627_17806\n",
+       "4818957       27965287_35726\n",
+       "8697796       9483753_328724\n",
+       "1049389      9446784_5019858\n",
+       "8364678      9467133_2925218\n",
+       "5119903    28880812_10305215\n",
        "Name: combined_id_catalog, dtype: object"
       ]
      },
-     "execution_count": 59,
+     "execution_count": 24,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -578,6 +709,187 @@
     "These pages are reasonably full too: [5660435](https://eol.org/pages/5660435)."
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Final Catalog\n",
+    "\n",
+    "Our final EOL catalog is thus going to be the `treeoflife_id`s in `links_inner`.\n",
+    "\n",
+    "Finall question is just how does this look for our `train_small`?"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 34,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "<class 'pandas.core.frame.DataFrame'>\n",
+      "Index: 593764 entries, 0 to 918500\n",
+      "Data columns (total 12 columns):\n",
+      " #   Column          Non-Null Count   Dtype  \n",
+      "---  ------          --------------   -----  \n",
+      " 0   split           593764 non-null  object \n",
+      " 1   treeoflife_id   593764 non-null  object \n",
+      " 2   eol_content_id  593764 non-null  float64\n",
+      " 3   eol_page_id     593764 non-null  float64\n",
+      " 4   kingdom         569128 non-null  object \n",
+      " 5   phylum          569285 non-null  object \n",
+      " 6   class           567356 non-null  object \n",
+      " 7   order           566839 non-null  object \n",
+      " 8   family          565269 non-null  object \n",
+      " 9   genus           564605 non-null  object \n",
+      " 10  species         565399 non-null  object \n",
+      " 11  common          593764 non-null  object \n",
+      "dtypes: float64(2), object(10)\n",
+      "memory usage: 58.9+ MB\n"
+     ]
+    }
+   ],
+   "source": [
+    "cat_2 = pd.read_csv(TOL_FILEPATH + \"catalog.csv\", low_memory=False)\n",
+    "\n",
+    "# reduce to just train_small and all non-EOL entries\n",
+    "cat_2 = cat_2.loc[cat_2.split == \"train_small\"]\n",
+    "eol_cat_2 = cat_2.loc[cat_2.eol_content_id.notna()]\n",
+    "\n",
+    "eol_cat_2 = eol_cat_2[eol_cols]\n",
+    "eol_cat_2.info(show_counts=True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "So `train_small` has 593,764 images from EOL. \n",
+    "\n",
+    "Let's recast IDs as ints to match up with `links_inner` and check we have info for all of them."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 35,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "<class 'pandas.core.frame.DataFrame'>\n",
+      "Index: 593764 entries, 0 to 918500\n",
+      "Data columns (total 4 columns):\n",
+      " #   Column          Non-Null Count   Dtype \n",
+      "---  ------          --------------   ----- \n",
+      " 0   split           593764 non-null  object\n",
+      " 1   treeoflife_id   593764 non-null  object\n",
+      " 2   eol_content_id  593764 non-null  int64 \n",
+      " 3   eol_page_id     593764 non-null  int64 \n",
+      "dtypes: int64(2), object(2)\n",
+      "memory usage: 22.7+ MB\n"
+     ]
+    }
+   ],
+   "source": [
+    "eol_cat_2 = eol_cat_2.astype({\"eol_content_id\": \"int64\", \"eol_page_id\": \"int64\"})\n",
+    "eol_cat_2[eol_cols[:4]].info()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 36,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# make combined ID for catalog and add suffix to eol content and page IDs\n",
+    "eol_cat_2['combined_id_catalog'] = eol_cat_2['eol_content_id'].astype(str) + '_' + eol_cat_2['eol_page_id'].astype(str)\n",
+    "eol_cat_2.rename(columns={'eol_content_id': 'eol_content_id_catalog', 'eol_page_id': 'eol_page_id_catalog'}, inplace=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 40,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "<class 'pandas.core.frame.DataFrame'>\n",
+      "RangeIndex: 612614 entries, 0 to 612613\n",
+      "Data columns (total 25 columns):\n",
+      " #   Column                           Non-Null Count   Dtype \n",
+      "---  ------                           --------------   ----- \n",
+      " 0   split                            612614 non-null  object\n",
+      " 1   treeoflife_id                    612614 non-null  object\n",
+      " 2   eol_content_id_catalog           612614 non-null  int64 \n",
+      " 3   eol_page_id_catalog              612614 non-null  int64 \n",
+      " 4   kingdom                          584979 non-null  object\n",
+      " 5   phylum                           585140 non-null  object\n",
+      " 6   class                            583235 non-null  object\n",
+      " 7   order                            582668 non-null  object\n",
+      " 8   family                           581082 non-null  object\n",
+      " 9   genus                            580464 non-null  object\n",
+      " 10  species                          579943 non-null  object\n",
+      " 11  common                           612614 non-null  object\n",
+      " 12  combined_id_catalog_x            612614 non-null  object\n",
+      " 13  combined_id_catalog_y            612614 non-null  object\n",
+      " 14  filename_manifest                612614 non-null  object\n",
+      " 15  combined_id_manifest_redownload  612614 non-null  object\n",
+      " 16  medium_source_url                612614 non-null  object\n",
+      " 17  eol_full_size_copy_url           612614 non-null  object\n",
+      " 18  license_name                     612614 non-null  object\n",
+      " 19  copyright_owner                  554029 non-null  object\n",
+      " 20  expected_image_filename          612614 non-null  object\n",
+      " 21  source_0706                      612614 non-null  bool  \n",
+      " 22  source_0726                      612614 non-null  bool  \n",
+      " 23  source_1206                      612614 non-null  bool  \n",
+      " 24  combined_id_full_manifest        612614 non-null  object\n",
+      "dtypes: bool(3), int64(2), object(20)\n",
+      "memory usage: 104.6+ MB\n"
+     ]
+    }
+   ],
+   "source": [
+    "eol_ts_links = pd.merge(eol_cat_2,\n",
+    "                        links_inner[list(links_inner.columns)[18:]],\n",
+    "                        left_on = \"combined_id_catalog\",\n",
+    "                        right_on = \"combined_id_full_manifest\",\n",
+    "                        how = \"inner\")\n",
+    "eol_ts_links.info(show_counts = True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 41,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "590730"
+      ]
+     },
+     "execution_count": 41,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "eol_ts_links.treeoflife_id.nunique()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "So we are missing the metadata for 3,734 of these images."
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -587,7 +899,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 60,
+   "execution_count": 25,
    "metadata": {},
    "outputs": [
     {
@@ -624,7 +936,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 61,
+   "execution_count": 26,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -634,7 +946,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 62,
+   "execution_count": 27,
    "metadata": {},
    "outputs": [
     {
@@ -643,39 +955,40 @@
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
       "RangeIndex: 12552 entries, 0 to 12551\n",
-      "Data columns (total 28 columns):\n",
-      " #   Column                          Non-Null Count  Dtype \n",
-      "---  ------                          --------------  ----- \n",
-      " 0   rarespecies_id                  12552 non-null  object\n",
-      " 1   eol_content_id_rs               12552 non-null  int64 \n",
-      " 2   eol_page_id_rs                  12552 non-null  int64 \n",
-      " 3   kingdom                         12552 non-null  object\n",
-      " 4   phylum                          12552 non-null  object\n",
-      " 5   class                           12552 non-null  object\n",
-      " 6   order                           12552 non-null  object\n",
-      " 7   family                          12552 non-null  object\n",
-      " 8   genus                           12552 non-null  object\n",
-      " 9   species                         12552 non-null  object\n",
-      " 10  sciName                         12552 non-null  object\n",
-      " 11  combined_id_rs                  12552 non-null  object\n",
-      " 12  eol_content_id                  12552 non-null  int64 \n",
-      " 13  eol_page_id                     12552 non-null  int64 \n",
-      " 14  medium_source_url               12552 non-null  object\n",
-      " 15  eol_full_size_copy_url          12552 non-null  object\n",
-      " 16  license_name                    12552 non-null  object\n",
-      " 17  copyright_owner                 11171 non-null  object\n",
-      " 18  expected_image_filename         12552 non-null  object\n",
-      " 19  source_0706                     12552 non-null  bool  \n",
-      " 20  source_0726                     12552 non-null  bool  \n",
-      " 21  source_1206                     12552 non-null  bool  \n",
-      " 22  combined_id_manifest            12552 non-null  object\n",
-      " 23  md5                             12552 non-null  object\n",
-      " 24  combined_id_manifest_checksums  12552 non-null  object\n",
-      " 25  eol_content_id_cargo            12552 non-null  int64 \n",
-      " 26  eol_page_id_cargo               12552 non-null  int64 \n",
-      " 27  combined_id_cargo               12552 non-null  object\n",
-      "dtypes: bool(3), int64(6), object(19)\n",
-      "memory usage: 2.4+ MB\n"
+      "Data columns (total 29 columns):\n",
+      " #   Column                           Non-Null Count  Dtype \n",
+      "---  ------                           --------------  ----- \n",
+      " 0   rarespecies_id                   12552 non-null  object\n",
+      " 1   eol_content_id_rs                12552 non-null  int64 \n",
+      " 2   eol_page_id_rs                   12552 non-null  int64 \n",
+      " 3   kingdom                          12552 non-null  object\n",
+      " 4   phylum                           12552 non-null  object\n",
+      " 5   class                            12552 non-null  object\n",
+      " 6   order                            12552 non-null  object\n",
+      " 7   family                           12552 non-null  object\n",
+      " 8   genus                            12552 non-null  object\n",
+      " 9   species                          12552 non-null  object\n",
+      " 10  sciName                          12552 non-null  object\n",
+      " 11  combined_id_rs                   12552 non-null  object\n",
+      " 12  filename_archive                 12552 non-null  object\n",
+      " 13  md5_archive                      12552 non-null  object\n",
+      " 14  combined_id_archive              12552 non-null  object\n",
+      " 15  filename_manifest                12552 non-null  object\n",
+      " 16  md5_combined_manifest            12552 non-null  object\n",
+      " 17  combined_id_manifest_redownload  12552 non-null  object\n",
+      " 18  eol_content_id                   12552 non-null  int64 \n",
+      " 19  eol_page_id                      12552 non-null  int64 \n",
+      " 20  medium_source_url                12552 non-null  object\n",
+      " 21  eol_full_size_copy_url           12552 non-null  object\n",
+      " 22  license_name                     12552 non-null  object\n",
+      " 23  copyright_owner                  11171 non-null  object\n",
+      " 24  expected_image_filename          12552 non-null  object\n",
+      " 25  source_0706                      12552 non-null  bool  \n",
+      " 26  source_0726                      12552 non-null  bool  \n",
+      " 27  source_1206                      12552 non-null  bool  \n",
+      " 28  combined_id_full_manifest        12552 non-null  object\n",
+      "dtypes: bool(3), int64(4), object(22)\n",
+      "memory usage: 2.5+ MB\n"
      ]
     }
    ],
@@ -683,51 +996,52 @@
     "rs_links = pd.merge(rs_catalog,\n",
     "                    links_manifest_cargo,\n",
     "                    left_on = \"combined_id_rs\",\n",
-    "                    right_on = \"combined_id_cargo\",\n",
+    "                    right_on = \"combined_id_archive\", # looking at archive\n",
     "                    how = \"inner\")\n",
     "rs_links.info(show_counts = True)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 63,
+   "execution_count": 28,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "rarespecies_id                    11826\n",
-       "eol_content_id_rs                 11826\n",
-       "eol_page_id_rs                      400\n",
-       "kingdom                               1\n",
-       "phylum                                5\n",
-       "class                                15\n",
-       "order                                85\n",
-       "family                              202\n",
-       "genus                               316\n",
-       "species                             385\n",
-       "sciName                             400\n",
-       "combined_id_rs                    11826\n",
-       "eol_content_id                    12221\n",
-       "eol_page_id                         447\n",
-       "medium_source_url                 12056\n",
-       "eol_full_size_copy_url            12119\n",
-       "license_name                         15\n",
-       "copyright_owner                    3724\n",
-       "expected_image_filename           12221\n",
-       "source_0706                           2\n",
-       "source_0726                           2\n",
-       "source_1206                           2\n",
-       "combined_id_manifest              12221\n",
-       "md5                               11663\n",
-       "combined_id_manifest_checksums    12221\n",
-       "eol_content_id_cargo              11826\n",
-       "eol_page_id_cargo                   400\n",
-       "combined_id_cargo                 11826\n",
+       "rarespecies_id                     11826\n",
+       "eol_content_id_rs                  11826\n",
+       "eol_page_id_rs                       400\n",
+       "kingdom                                1\n",
+       "phylum                                 5\n",
+       "class                                 15\n",
+       "order                                 85\n",
+       "family                               202\n",
+       "genus                                316\n",
+       "species                              385\n",
+       "sciName                              400\n",
+       "combined_id_rs                     11826\n",
+       "filename_archive                   11826\n",
+       "md5_archive                        11663\n",
+       "combined_id_archive                11826\n",
+       "filename_manifest                  12221\n",
+       "md5_combined_manifest              11663\n",
+       "combined_id_manifest_redownload    12221\n",
+       "eol_content_id                     12221\n",
+       "eol_page_id                          447\n",
+       "medium_source_url                  12056\n",
+       "eol_full_size_copy_url             12119\n",
+       "license_name                          15\n",
+       "copyright_owner                     3724\n",
+       "expected_image_filename            12221\n",
+       "source_0706                            2\n",
+       "source_0726                            2\n",
+       "source_1206                            2\n",
+       "combined_id_full_manifest          12221\n",
        "dtype: int64"
       ]
      },
-     "execution_count": 63,
+     "execution_count": 28,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -745,7 +1059,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 64,
+   "execution_count": 29,
    "metadata": {},
    "outputs": [
     {
@@ -754,7 +1068,7 @@
        "11826"
       ]
      },
-     "execution_count": 64,
+     "execution_count": 29,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -766,7 +1080,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 65,
+   "execution_count": 30,
    "metadata": {},
    "outputs": [
     {
@@ -778,7 +1092,7 @@
        " '29975068_45509269']"
       ]
      },
-     "execution_count": 65,
+     "execution_count": 30,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -789,20 +1103,20 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 66,
+   "execution_count": 31,
    "metadata": {},
    "outputs": [
     {
      "data": {
       "text/plain": [
-       "4010      20184952_311570\n",
-       "8096      29473123_323892\n",
-       "5217    29580624_45514697\n",
-       "890     21629967_46559745\n",
+       "1473     29538601_46580048\n",
+       "11127     29447084_1019571\n",
+       "6578       20398296_328017\n",
+       "11583     29772112_1049135\n",
        "Name: combined_id_rs, dtype: object"
       ]
      },
-     "execution_count": 66,
+     "execution_count": 31,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -813,7 +1127,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 67,
+   "execution_count": 32,
    "metadata": {},
    "outputs": [
     {
@@ -849,7 +1163,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 68,
+   "execution_count": 33,
    "metadata": {},
    "outputs": [
     {
@@ -858,38 +1172,39 @@
      "text": [
       "<class 'pandas.core.frame.DataFrame'>\n",
       "RangeIndex: 0 entries\n",
-      "Data columns (total 28 columns):\n",
-      " #   Column                          Non-Null Count  Dtype \n",
-      "---  ------                          --------------  ----- \n",
-      " 0   rarespecies_id                  0 non-null      object\n",
-      " 1   eol_content_id_rs               0 non-null      int64 \n",
-      " 2   eol_page_id_rs                  0 non-null      int64 \n",
-      " 3   kingdom                         0 non-null      object\n",
-      " 4   phylum                          0 non-null      object\n",
-      " 5   class                           0 non-null      object\n",
-      " 6   order                           0 non-null      object\n",
-      " 7   family                          0 non-null      object\n",
-      " 8   genus                           0 non-null      object\n",
-      " 9   species                         0 non-null      object\n",
-      " 10  sciName                         0 non-null      object\n",
-      " 11  combined_id_rs                  0 non-null      object\n",
-      " 12  eol_content_id                  0 non-null      int64 \n",
-      " 13  eol_page_id                     0 non-null      int64 \n",
-      " 14  medium_source_url               0 non-null      object\n",
-      " 15  eol_full_size_copy_url          0 non-null      object\n",
-      " 16  license_name                    0 non-null      object\n",
-      " 17  copyright_owner                 0 non-null      object\n",
-      " 18  expected_image_filename         0 non-null      object\n",
-      " 19  source_0706                     0 non-null      bool  \n",
-      " 20  source_0726                     0 non-null      bool  \n",
-      " 21  source_1206                     0 non-null      bool  \n",
-      " 22  combined_id_manifest            0 non-null      object\n",
-      " 23  md5                             0 non-null      object\n",
-      " 24  combined_id_manifest_checksums  0 non-null      object\n",
-      " 25  eol_content_id_cargo            0 non-null      int64 \n",
-      " 26  eol_page_id_cargo               0 non-null      int64 \n",
-      " 27  combined_id_cargo               0 non-null      object\n",
-      "dtypes: bool(3), int64(6), object(19)\n",
+      "Data columns (total 29 columns):\n",
+      " #   Column                           Non-Null Count  Dtype \n",
+      "---  ------                           --------------  ----- \n",
+      " 0   rarespecies_id                   0 non-null      object\n",
+      " 1   eol_content_id_rs                0 non-null      int64 \n",
+      " 2   eol_page_id_rs                   0 non-null      int64 \n",
+      " 3   kingdom                          0 non-null      object\n",
+      " 4   phylum                           0 non-null      object\n",
+      " 5   class                            0 non-null      object\n",
+      " 6   order                            0 non-null      object\n",
+      " 7   family                           0 non-null      object\n",
+      " 8   genus                            0 non-null      object\n",
+      " 9   species                          0 non-null      object\n",
+      " 10  sciName                          0 non-null      object\n",
+      " 11  combined_id_rs                   0 non-null      object\n",
+      " 12  filename_archive                 0 non-null      object\n",
+      " 13  md5_archive                      0 non-null      object\n",
+      " 14  combined_id_archive              0 non-null      object\n",
+      " 15  filename_manifest                0 non-null      object\n",
+      " 16  md5_combined_manifest            0 non-null      object\n",
+      " 17  combined_id_manifest_redownload  0 non-null      object\n",
+      " 18  eol_content_id                   0 non-null      int64 \n",
+      " 19  eol_page_id                      0 non-null      int64 \n",
+      " 20  medium_source_url                0 non-null      object\n",
+      " 21  eol_full_size_copy_url           0 non-null      object\n",
+      " 22  license_name                     0 non-null      object\n",
+      " 23  copyright_owner                  0 non-null      object\n",
+      " 24  expected_image_filename          0 non-null      object\n",
+      " 25  source_0706                      0 non-null      bool  \n",
+      " 26  source_0726                      0 non-null      bool  \n",
+      " 27  source_1206                      0 non-null      bool  \n",
+      " 28  combined_id_full_manifest        0 non-null      object\n",
+      "dtypes: bool(3), int64(4), object(22)\n",
       "memory usage: 132.0+ bytes\n"
      ]
     }
@@ -898,7 +1213,7 @@
     "links_rs_mismatch = pd.merge(mismatched_rs,\n",
     "                             links_mismatch,\n",
     "                             left_on = \"combined_id_rs\",\n",
-    "                             right_on = \"combined_id_manifest_checksums\",\n",
+    "                             right_on = \"combined_id_full_manifest\",\n",
     "                             how = \"inner\")\n",
     "links_rs_mismatch.info(show_counts = True)"
    ]

eol_realign/notebooks/links_align_reduced.py

@@ -16,12 +16,33 @@
 # %%
 import pandas as pd
 
+# %%
+TOL_FILEPATH = "../../data/"
+
 # %% [markdown]
 # ### Read in Links CSV files
+#
+# Full cargo archives have some more content than the [original `links_inner` file](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_inner.csv) and the [original `links_manifest_cargo_on_md5.csv`](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_manifest_cargo_on_md5.csv).
+#
+# From Matt:
+# > It represents a series of inner merges:
+# > 1. The list of images in the EOL cargo archive (used to create the webdatasets) with the EOL portion of `catalog.csv`
+# >  - Merged on a combined identifier of `eol_content_id` and `eol_page_id`.
+# > Result from 1) with the list of images redownloaded from all 3 media manifests we touched around the original download time (identified [here](https://huggingface.co/datasets/imageomics/ToL-EDA/discussions/5#6584808d92eba5ad69b8379f)).
+# >  - Merged on MD5
+# >
+# > The result has 6,219,674 unique `treeoflife_id` entries and 6,146,917 unique `md5` entries. It has the combined `eol_content_id` and `eol_page_id` identifiers used at each stage:
+# - `combined_id_catalog`: represents what's in the original catalog 
+# - `combined_id_manifest_full_manifest`: represents what's in the combined manifest from all 3 dates 
+# > It also has an indicator of which of the three media manifest files each entry was present in (one or more) in columns representing retreival date (eg., `source_0706`: indicates if it was present in the manifest retrieved on July 6, 2023). It also has the copyright owner and license information as contained in the media manifest files.
+# >
+# > Still needs to be run through the owner name finding code. ([owner_match](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/71ad5b49226a3a49634931005eeb340b3c96db13/scripts/match_owners.py) --note: this version uses inner merges, so will remove entries without media info. Probably should adjust this.)
+#
+# For reference on the numbers, there are 6,250,420 EOL images in TreeOfLife-10M (the catalog). Comparing with `links_inner` previously (from [`links_inner.csv`](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_inner.csv)) we were missing 31,376 images, 29 of which were identified in [`links_manifest_cargo_on_md5.csv`](https://huggingface.co/datasets/imageomics/ToL-EDA/blob/fa08e0f6692a013e4e4f5ca30dd0e8883c0021d8/eol_realign/data/links_manifest_cargo_on_md5.csv).
 
 # %%
-links_inner = pd.read_csv("../data/links_inner.csv", low_memory=False)
-links_manifest_cargo = pd.read_csv("../data/links_manifest_cargo_on_md5.csv", low_memory=False)
+links_inner = pd.read_csv("../data/eol-cargo-archive_catalog_combined-manifest-checksums_links.csv", low_memory=False)
+links_manifest_cargo = pd.read_csv("../data/eol-cargo-archive_combined-manifest-checksums_links.csv", low_memory=False)
 
 # %%
 links_inner.info(show_counts=True)
@@ -29,18 +50,31 @@ links_inner.info(show_counts=True)
 # %%
 links_manifest_cargo.info(show_counts=True)
 
+# %% [markdown]
+# These both have about 600-700 more entries than the earlier files.
+
 # %%
-links_mismatch = links_manifest_cargo.loc[links_manifest_cargo["combined_id_cargo"] != links_manifest_cargo["combined_id_manifest_checksums"]]
+# These should be the same
+links_manifest_cargo.loc[links_manifest_cargo["combined_id_manifest_redownload"] != links_manifest_cargo["combined_id_full_manifest"]]
+
+# %%
+# archive has replaced cargo
+# full manifest or manifest redownload to replace manifest checksums -- they are equal
+links_mismatch = links_manifest_cargo.loc[links_manifest_cargo["combined_id_archive"] != links_manifest_cargo["combined_id_full_manifest"]]
 links_mismatch.info(show_counts = True)
 
+# %% [markdown]
+# There are 42 more mismatched entries. Only 13 additional unique images (MD5s).
+
 # %%
-links_mismatch[["md5", "combined_id_manifest_checksums", "combined_id_cargo"]].nunique()
+links_mismatch[["md5_archive", "combined_id_full_manifest", "combined_id_archive"]].nunique()
 
 # %%
-links_mismatch.to_csv("../data/links_cargo_manifest_IDmismatch.csv", index = False)
+# This is archive manifest mismatch
+links_mismatch.to_csv("../data/links_archive_manifest_IDmismatch.csv", index = False)
 
 # %%
-catalog = pd.read_csv("../../data/catalog.csv", low_memory=False)
+catalog = pd.read_csv(TOL_FILEPATH + "catalog.csv", low_memory=False)
 
 # %%
 # remove train_small and all non-EOL entries
@@ -74,11 +108,14 @@ eol_catalog.rename(columns={'eol_content_id': 'eol_content_id_catalog', 'eol_pag
 matched_catalog_ids = list(links_inner.combined_id_catalog.unique())
 len(matched_catalog_ids)
 
+# %% [markdown]
+# Here we go, we got an additional 630 matched IDs!
+
 # %%
 matched_catalog_ids[:5]
 
 # %% [markdown]
-# So we have about 30K unmatched, let's find those in the catalog and then we'll try to merge with our mismatched cargo & manifest.
+# We still have about 30K unmatched (as shown below, it's down to 30,746 from 31,376), let's find those in the catalog and then we'll try to merge with our mismatched cargo & manifest.
 
 # %%
 eol_catalog['combined_id_catalog'].sample(5)
@@ -91,24 +128,76 @@ mismatched_catalog.info(show_counts = True)
 mismatched_catalog.nunique()
 
 # %% [markdown]
-# ## Merge with Mismatched Cargo-Manifest
+# Seems one class got resolved (equals more orders and down).
+
+# %% [markdown]
+# ## Merge with Mismatched CargoArchive-Manifest
 #
-# let's merge with `links_mismatch` to see if the mismatched cargo and manifest combined IDs can be linked up.
+# let's merge with `links_mismatch` to see if the mismatched archive and manifest combined IDs can be linked up.
 
 # %%
+# replace manifest checksums with full manifest as above
 links_catalog_mismatch = pd.merge(mismatched_catalog,
                                   links_mismatch,
                                   left_on = "combined_id_catalog",
-                                  right_on = "combined_id_manifest_checksums",
+                                  right_on = "combined_id_full_manifest",
                                   how = "inner")
 links_catalog_mismatch.info(show_counts = True)
 
+# %% [markdown]
+# Now there's no gain from the mismatch file, but that was only 29 images, so we've captured them from the archive (plus more!).
+
 # %%
 mismatched_catalog.combined_id_catalog.sample(7)
 
 # %% [markdown]
 # These pages are reasonably full too: [5660435](https://eol.org/pages/5660435).
 
+# %% [markdown]
+# ### Final Catalog
+#
+# Our final EOL catalog is thus going to be the `treeoflife_id`s in `links_inner`.
+#
+# Finall question is just how does this look for our `train_small`?
+
+# %%
+cat_2 = pd.read_csv(TOL_FILEPATH + "catalog.csv", low_memory=False)
+
+# reduce to just train_small and all non-EOL entries
+cat_2 = cat_2.loc[cat_2.split == "train_small"]
+eol_cat_2 = cat_2.loc[cat_2.eol_content_id.notna()]
+
+eol_cat_2 = eol_cat_2[eol_cols]
+eol_cat_2.info(show_counts=True)
+
+# %% [markdown]
+# So `train_small` has 593,764 images from EOL. 
+#
+# Let's recast IDs as ints to match up with `links_inner` and check we have info for all of them.
+
+# %%
+eol_cat_2 = eol_cat_2.astype({"eol_content_id": "int64", "eol_page_id": "int64"})
+eol_cat_2[eol_cols[:4]].info()
+
+# %%
+# make combined ID for catalog and add suffix to eol content and page IDs
+eol_cat_2['combined_id_catalog'] = eol_cat_2['eol_content_id'].astype(str) + '_' + eol_cat_2['eol_page_id'].astype(str)
+eol_cat_2.rename(columns={'eol_content_id': 'eol_content_id_catalog', 'eol_page_id': 'eol_page_id_catalog'}, inplace=True)
+
+# %%
+eol_ts_links = pd.merge(eol_cat_2,
+                        links_inner[list(links_inner.columns)[18:]],
+                        left_on = "combined_id_catalog",
+                        right_on = "combined_id_full_manifest",
+                        how = "inner")
+eol_ts_links.info(show_counts = True)
+
+# %%
+eol_ts_links.treeoflife_id.nunique()
+
+# %% [markdown]
+# So we are missing the metadata for 3,734 of these images.
+
 # %% [markdown]
 # ## Check on Rare Species Catalog
 
@@ -126,7 +215,7 @@ rs_catalog.rename(columns={'eol_content_id': 'eol_content_id_rs', 'eol_page_id':
 rs_links = pd.merge(rs_catalog,
                     links_manifest_cargo,
                     left_on = "combined_id_rs",
-                    right_on = "combined_id_cargo",
+                    right_on = "combined_id_archive", # looking at archive
                     how = "inner")
 rs_links.info(show_counts = True)
 
@@ -154,7 +243,7 @@ mismatched_rs.info(show_counts = True)
 links_rs_mismatch = pd.merge(mismatched_rs,
                              links_mismatch,
                              left_on = "combined_id_rs",
-                             right_on = "combined_id_manifest_checksums",
+                             right_on = "combined_id_full_manifest",
                              how = "inner")
 links_rs_mismatch.info(show_counts = True)