How to prove every Purview glossary term arrived as a Fabric tag
The short answer
Keep your checked Purview export, export Fabric's tags, items and domains, and record which Fabric item each Purview data asset became. One Python and DuckDB script runs seven checks that each published term became a tag on the right items. The script writes one sign-off sheet per governance domain: counts, differences, terms to check by hand and a signature line.
Why a steward needs a reconciliation to sign
Microsoft announced on November 12, 2026 that Purview Data Governance is retiring: “Support for Purview Data Governance will end Nov. 12, 2027” (Microsoft Security blog, read 2026-11-21). As of 2026-11-21, Microsoft's transition guide (read 2026-11-21) maps the glossary to tags, descriptions and endorsement in the OneLake catalog. A Fabric tag is a name: it holds no definition, owner or parent (Tags in Microsoft Fabric, read 2026-11-21).
So a glossary that moves to Fabric arrives as tag names on items, and the rest of each term stays behind. A steward who signs that a domain arrived needs two things in writing: which terms became tags on the right items, and which parts of each term did not move. This guide compares your Purview export with a Fabric export and writes both, one sheet per governance domain.
What you will have at the end
- Seven checks, each one PASS or FAIL, that every published term became a tag of the same name on the Fabric items its data assets became.
- One sign-off sheet per governance domain: the counts on both sides, every difference in the domain and every difference that belongs to no one domain, what a tag does not carry, a sample of terms to check by hand, the owners of the moved terms and a signature line.
- Two CSV files:
differences.csv(one row per difference) andmoved-terms.csv(one row per published term, with its tag and the items it should be on).
What you need
- The Purview export, saved and checked as the guide How to export Purview glossary terms with their relationships shows. The reconciliation stops if a term id appears twice, a page was not saved, or a term belongs to a domain that is missing from the domain names, so run the export check first. Save the domain names as well (step 4 of that guide's commands): the reconciliation matches domains by name.
- A Fabric export of tags, items and domains, saved with the commands in step 1.
- An asset map: a CSV with the columns
purview_asset_idandfabric_item_id, one row for each Purview data asset and the Fabric item that asset became. Purview and Fabric give the same table different ids, so only your migration's own record can join them.
Step 1. Save what landed in Fabric
The commands read three Fabric admin APIs: List Tags, List Items, which returns each item's tags (as of 2026-11-21, List Items is a preview API), and List Domains (all read 2026-11-21). A Fabric administrator signed in as a user needs the Tenant.Read.All scope. The commands below sign in as a service principal instead. That service principal needs no API permissions on its app registration: a Fabric administrator turns on the admin API settings for service principals and adds the app to the security group named there. List Items allows 10 requests a minute and 200 an hour for each tenant, so the commands pause between its pages. Microsoft describes the admin API settings for service principals (read 2026-11-21). The token request follows Microsoft's client credentials flow (read 2026-11-21).
We did not run these commands. We have no Fabric tenant to run them against, so they are written from Microsoft's reference pages, and the runs further down use generated records in the same shape. They need curl and jq.
# Not run by us: written from Microsoft's Fabric REST API reference (read 2026-11-21).
# 1. A token for a service principal that a Fabric administrator allowed to call admin APIs (no API permissions on the app)
token() {
curl -s -X POST "https://login.microsoftonline.com/$TENANT_ID/oauth2/v2.0/token" \
-d grant_type=client_credentials -d client_id="$CLIENT_ID" \
--data-urlencode client_secret="$CLIENT_SECRET" -d scope=https://api.fabric.microsoft.com/.default \
| jq -r .access_token
}
fetch() { # fetch URL FILE: save the answer; on 401 get a new token and try once more
code=$(curl -s -o "$2" -w '%{http_code}' -H "Authorization: Bearer $TOKEN" "$1")
if [ "$code" = 401 ]; then
TOKEN=$(token)
code=$(curl -s -o "$2" -w '%{http_code}' -H "Authorization: Bearer $TOKEN" "$1")
fi
[ "$code" = 200 ] || echo "HTTP $code: $2 holds an error, fetch $1 again" >&2
}
pages() { # pages URL FOLDER PAUSE: every page, following continuationUri, one file per page
url="$1"; n=1
while [ -n "$url" ]; do
f=$(printf "%s/page-%03d.json" "$2" "$n")
fetch "$url" "$f"
url=$(jq -r '.continuationUri // empty' "$f"); n=$((n + 1))
[ -n "$url" ] && sleep "$3"
done
}
TOKEN=$(token)
OUT="fabric-$(date +%Y%m%d-%H%M%S)" # a new, empty folder for every export
mkdir -p "$OUT/tags" "$OUT/items" "$OUT/domains"
pages "https://api.fabric.microsoft.com/v1/admin/tags" "$OUT/tags" 3 # 25 requests a minute
pages "https://api.fabric.microsoft.com/v1/admin/domains" "$OUT/domains" 3 # 25 requests a minute
pages "https://api.fabric.microsoft.com/v1/admin/items" "$OUT/items" 7 # 10 requests a minute, 200 an hour
echo "Saved to $OUT"
Each run writes a new folder named fabric- with the date and time. When Fabric answers 401 because the token expired, the commands get a new token and try the request once more. If a request fails for another reason, for example 429 when the tenant's rate limit is used up, the commands print the file's name and stop reading that API's pages, and the reconciliation stops on that file. Wait a few minutes and run the commands again: each run writes a new folder.
Step 2. Run the reconciliation
Download reconcile_catalog.py, install DuckDB (pip install duckdb) and run the script with the three inputs. The script reads the files with Python, loads them into DuckDB in memory and runs seven checks in SQL. Nothing is sent anywhere.
python3 reconcile_catalog.py --purview export-20261121-093000 --fabric fabric-20261121-101500 --asset-map asset-map.csv
| Check | What a failure means |
|---|---|
| Every Purview domain has a Fabric domain of the same name | The terms of that domain have no domain to land in. The script matches domain names ignoring case and spacing; a name that matches more than one Fabric domain also fails, because the terms could land in any of those Fabric domains. |
| Every published term has a Fabric tag of the same name | The term did not arrive. A tag counts if it sits in the term's Fabric domain or at the tenant level. |
| No tag differs from its term only in case or spacing | The tag arrived under another spelling, so searches for the term's exact name miss the tag. |
| Every data asset linked to a published term is in the asset map | Nobody recorded which Fabric item the data asset became, so the link cannot be checked. |
| Every Fabric item in the asset map is in the Fabric export | The asset map points to an item that does not exist, or the Fabric export missed the item. The difference appears on the sheet of every domain whose published terms link to that item. An asset map row that no published term uses appears on every sheet, as a difference outside any one domain. |
| Every term-to-asset link is a tag on the mapped Fabric item | The tag exists, but an item that should carry the tag does not. A tenant tag or a domain tag with the term's exact name both count. |
| Every tag in a mapped Fabric domain comes from a Purview term | A tag in the domain has no published term behind it, for example a test tag or a term that was meant to stay in draft. |
Only published terms are expected in Fabric. Draft and expired terms are counted on the sheet as not moved. To move them too, change MOVE_STATUSES at the top of the script.
Run 1. A generated migration, clean
The script writes a generated Purview export, a generated Fabric landing and an asset map when you ask for them: three invented domains with 20 terms each, in the shapes Microsoft's APIs return, with a fixed seed so every run gives the same records.
python3 reconcile_catalog.py --make-demo demo
Wrote a generated Purview export and Fabric landing to demo: 60 terms in 3 Purview domains, 51 tags, 24 items, 24 asset map rows (seed 278)
python3 reconcile_catalog.py --purview demo/purview --fabric demo/fabric --asset-map demo/asset-map.csv
Python 3.14.8 duckdb 1.5.6 Purview: 60 terms (51 published, so expected in Fabric), 120 links to data assets, 3 domains Fabric: 51 tags, 24 items, 3 domains; asset map: 24 rows Checks PASS Every Purview domain has a Fabric domain of the same name PASS Every published term has a Fabric tag of the same name PASS No tag differs from its term only in case or spacing PASS Every data asset linked to a published term is in the asset map PASS Every Fabric item in the asset map is in the Fabric export PASS Every term-to-asset link is a tag on the mapped Fabric item PASS Every tag in a mapped Fabric domain comes from a Purview term Wrote differences.csv (0 rows), moved-terms.csv (51 rows) and 3 sign-off sheets to .: signoff-finance.md, signoff-sales.md, signoff-supply-chain.md RESULT: no differences. Each domain's steward can check the sample and sign.
The sign-off sheet for one domain
This is signoff-finance.md from run 1, as the script wrote the sheet:
# Sign-off: Finance domain Purview governance domain Finance (a8682767-6848-4a9d-9a7d-025615af9991) moved to Fabric domain Finance (a509fd64-3ee8-47fb-81d4-3ba7da5b0875). | What | Purview | Fabric | |---|---|---| | Published terms / tags of the same name | 17 | 17 | | Term-to-asset links / tags on the mapped items | 34 | 34 | Differences found by the check: 0 (none). Not carried by a Fabric tag (keep in the archive or another home): - 17 term descriptions, 5 terms with acronyms, 2 parent links - draft or expired terms, not moved: 3 Terms to check by hand (10, chosen with seed 278): open each tag in Fabric and confirm the tag exists and sits on the items listed for the term in moved-terms.csv. - [ ] Accounts Receivable - [ ] Accrual - [ ] Cost Center - [ ] Days Sales Outstanding - [ ] Deferred Revenue - [ ] Fiscal Period - [ ] Net Revenue - [ ] Operating Expense - [ ] Revenue Recognition Date - [ ] Write-off Owners of the moved terms (Microsoft Entra object ids, number of terms): - fcc147c4-a265-4038-bbf4-84c70a7dbc3c: 7 - 393383a2-5c49-4ab6-b8fc-4bd7ae5e84ba: 5 - 472d4fc8-7f68-4538-8b24-d1e8e8275436: 5 Signed means: the counts above were compared, every difference is fixed or accepted in writing, and the terms above were checked by hand. Steward: ______________________ Date: ____________ Signature: ______________________
Run 2. The same migration with seven planted differences
python3 reconcile_catalog.py --make-demo demo-planted --plant-errors
Wrote a generated Purview export and Fabric landing to demo-planted: 60 terms in 4 Purview domains, 51 tags, 24 items, 25 asset map rows (seed 278) Planted one difference for each check: ASSET_NOT_MAPPED, DOMAIN_NOT_MAPPED, EXTRA_TAG, ITEM_MISSING, LINK_NOT_TAGGED, TAG_RENAMED, TERM_WITHOUT_TAG
python3 reconcile_catalog.py --purview demo-planted/purview --fabric demo-planted/fabric --asset-map demo-planted/asset-map.csv --tag planted
Python 3.14.8 duckdb 1.5.6 Purview: 60 terms (51 published, so expected in Fabric), 121 links to data assets, 4 domains Fabric: 51 tags, 24 items, 3 domains; asset map: 25 rows Checks FAIL Every Purview domain has a Fabric domain of the same name (1) FAIL Every published term has a Fabric tag of the same name (1) FAIL No tag differs from its term only in case or spacing (1) FAIL Every data asset linked to a published term is in the asset map (1) FAIL Every Fabric item in the asset map is in the Fabric export (1) FAIL Every term-to-asset link is a tag on the mapped Fabric item (1) FAIL Every tag in a mapped Fabric domain comes from a Purview term (1) Differences DOMAIN_NOT_MAPPED [Human Resources] no Fabric domain named "Human Resources" TERM_WITHOUT_TAG [Sales] Win Rate: no tag named "Win Rate" in the domain or the tenant TAG_RENAMED [Supply Chain] Lead Time: the tag is named "lead time" ASSET_NOT_MAPPED [Sales] Quote: data asset c9de0790-ebd5-4925-a714-9b57e9dd9844 has no row in the asset map ITEM_MISSING asset map row 13d97393-b934-4278-aa90-d6e2aaf2f64a -> b95dc197-c233-483f-9467-98c99b9c31d8: no Fabric item b95dc197-c233-483f-9467-98c99b9c31d8 in the export LINK_NOT_TAGGED [Finance] Gross Margin: Fabric item finance_table_7 does not carry the tag EXTRA_TAG [Finance] tag "Test tag" has no published term of that name Wrote differences-planted.csv (7 rows), moved-terms-planted.csv (51 rows) and 4 sign-off sheets to .: signoff-finance-planted.md, signoff-human-resources-planted.md, signoff-sales-planted.md, signoff-supply-chain-planted.md RESULT: 7 differences. Fix them in Fabric, or accept each one in writing, before a steward signs.
| Check | The difference planted in run 2 |
|---|---|
| Domains mapped | A fourth Purview domain, Human Resources, with no Fabric domain. |
| Every published term has a tag | The Win Rate tag was never created. |
| No renamed tags | The Lead Time tag was created as “lead time”. |
| Every linked asset is mapped | Quote is linked to a data asset the asset map does not hold. |
| Every mapped item exists | One asset map row points to an item that is not in the Fabric export. |
| Every link is a tag on its item | One item lost the Gross Margin tag. |
| No extra tags | A tag named “Test tag” left in the Finance domain. |
What to do with each difference
- Fix it in Fabric when the move went wrong: create the missing tag, rename the tag, apply the tag to the item, or delete the test tag. Then export Fabric again and re-run.
- Fix the asset map when the record of which item each data asset became is wrong or incomplete.
- Accept it in writing when the difference is a decision, for example a domain that is being retired. The sheet says that signing means every difference is fixed or accepted in writing.
Run the reconciliation on your own data
To run the reconciliation on your own migration, save the checked Purview export, the Fabric export from step 1 and your asset map, then run python3 reconcile_catalog.py with --purview, --fabric and --asset-map pointing at those three inputs. You do not edit the script. The script reads saved files and changes nothing in Purview or Fabric. Add --fail-on-differences to make it exit with code 1 when a check fails, and --out <folder> to write the sheets somewhere other than the script's folder.
We tested a run on a reader's own data: a second generated migration with other domain and term names, upper-case ids, empty fields, a tenant-level tag, two pages of items and at least one planted difference for each check. All seven checks failed as planted, and nothing else was flagged.
Download reconcile_catalog.py · Get the full guide (PDF)
Results from the runs above: moved-terms.csv and signoff-finance.md from run 1, and differences-planted.csv from run 2.
Get the full guide as a PDF
We'll email you a link to it. Please use your work email.
Thanks, . We're sending the link to now. If it isn't there in a few minutes, check your spam or junk folder, or email info@datameaning.com.
Limits
- The runs on this page use generated records, not a real Purview or Fabric tenant, and the export commands in step 1 were not run. The field names come from Microsoft's reference pages and API specifications, read on 2026-11-21.
- The reconciliation checks glossary terms that became Fabric tags. A term that moved to another catalog needs that catalog's export, laid out in the same three folders, before the script can compare it.
- A Fabric tag holds no owner, so the sheet lists the owners recorded in Purview as the people who sign; the check cannot confirm that those people still own the data in Fabric.
- The sample of terms to check by hand is chosen with a fixed seed, so the same export always gives the same sample. A steward who wants a different sample changes
SAMPLE_SEED. - Dates and limits here are as of 2026-11-21. This guide is about Purview Data Governance, whose support ends on November 12, 2027; after that date the guide describes history.
Tested on
Tested on
- Platform
- DuckDB 1.5.6, run from Python 3.14.8
- Edition
- Open-source DuckDB library (
pip install duckdb) - Date
- 2026-11-21
- Records
- Generated: an invented glossary of 60 terms in three domains and its invented Fabric landing, fixed seed 278, in the shapes of the Unified Catalog API version 2026-03-20-preview and the Fabric admin API (Fabric API specification at commit ec85f05, read 2026-11-21). Not a run against a Purview or Fabric tenant.
- Code
- reconcile_catalog.py
- Results
- moved-terms.csv, signoff-finance.md, differences-planted.csv
When this is not enough
This reconciliation tells a steward whether the glossary arrived in Fabric as tags. The harder questions sit around the glossary: where each kind of content should land in the first place, what happens to descriptions, parents and data quality rules that a tag cannot hold, which reports still read the Purview catalog, and when Purview can be switched off.
Moving your whole Purview governance catalog, not only the glossary?
Purview to OneLake Migration is in early access. The offering starts with the four-day Catalog Transition Assessment: one fixed fee of $12,000, and no commitment to anything else. You get an inventory of every object in your Purview governance catalog and a decision for each object. AI does the first pass, and your stewards sign every domain.