tdsql-mcp 1.4.4__tar.gz → 1.4.5__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (59) hide show
  1. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/PKG-INFO +1 -1
  2. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/pyproject.toml +2 -1
  3. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/aggregate-functions.md +16 -14
  4. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/association-analysis.md +35 -9
  5. tdsql_mcp-1.4.5/skills/teradata-sql-analytics/syntax/authorization-objects.md +259 -0
  6. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/bit-byte-functions.md +25 -15
  7. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/catalog-views.md +26 -17
  8. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/conditional.md +16 -8
  9. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/data-cleaning.md +33 -44
  10. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/data-exploration.md +34 -15
  11. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/data-prep.md +55 -36
  12. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/data-types-casting.md +8 -6
  13. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/date-time.md +25 -11
  14. tdsql_mcp-1.4.5/skills/teradata-sql-analytics/syntax/fit-transform-pattern.md +123 -0
  15. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/geospatial.md +14 -6
  16. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/guidelines.md +17 -14
  17. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/hypothesis-testing.md +23 -19
  18. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/index.md +7 -15
  19. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/json-functions.md +21 -12
  20. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/ml-functions.md +34 -24
  21. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/ml-patterns.md +24 -17
  22. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/model-evaluation.md +47 -36
  23. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/numeric-functions.md +13 -7
  24. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/object-store.md +39 -7
  25. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/open-table-format.md +75 -33
  26. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/path-analysis.md +23 -52
  27. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/query-tuning.md +19 -22
  28. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/sql-basics.md +38 -61
  29. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/string-functions.md +26 -18
  30. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/text-analytics.md +39 -77
  31. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-concepts.md +7 -0
  32. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-data-prep.md +9 -1
  33. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-diagnostics.md +15 -12
  34. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-dsp.md +13 -10
  35. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-estimation.md +22 -10
  36. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-forecasting.md +17 -17
  37. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-formula-rules.md +11 -13
  38. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/uaf-utility.md +24 -10
  39. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/utility-functions.md +17 -12
  40. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/vector-search.md +31 -17
  41. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/window-functions.md +25 -10
  42. tdsql_mcp-1.4.4/skills/teradata-sql-analytics/syntax/authorization-objects.md +0 -239
  43. tdsql_mcp-1.4.4/skills/teradata-sql-analytics/syntax/fit-transform-pattern.md +0 -94
  44. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/.claude-plugin/plugin.json +0 -0
  45. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/.env.example +0 -0
  46. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/.gitignore +0 -0
  47. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/CLAUDE.md +0 -0
  48. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/README.md +0 -0
  49. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/docs/architecture.md +0 -0
  50. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/requirements.txt +0 -0
  51. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/README.md +0 -0
  52. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/SKILL.md +0 -0
  53. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/ai-text-analytics.md +0 -0
  54. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/byom-model-loading.md +0 -0
  55. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/byom-scoring.md +0 -0
  56. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/embeddings.md +0 -0
  57. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/skills/teradata-sql-analytics/syntax/llm-providers.md +0 -0
  58. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/src/tdsql_mcp/__init__.py +0 -0
  59. {tdsql_mcp-1.4.4 → tdsql_mcp-1.4.5}/src/tdsql_mcp/server.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.5
2
2
  Name: tdsql-mcp
3
- Version: 1.4.4
3
+ Version: 1.4.5
4
4
  Summary: MCP server for Teradata Vantage — SQL execution and native analytics function reference for AI agents
5
5
  Project-URL: Homepage, https://github.com/ksturgeon-td/tdsql-mcp
6
6
  Project-URL: Repository, https://github.com/ksturgeon-td/tdsql-mcp
@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
4
4
 
5
5
  [project]
6
6
  name = "tdsql-mcp"
7
- version = "1.4.4"
7
+ version = "1.4.5"
8
8
  description = "MCP server for Teradata Vantage — SQL execution and native analytics function reference for AI agents"
9
9
  readme = "README.md"
10
10
  requires-python = ">=3.10"
@@ -41,3 +41,4 @@ tdsql-mcp = "tdsql_mcp.server:main"
41
41
 
42
42
  [tool.hatch.build.targets.wheel]
43
43
  packages = ["src/tdsql_mcp"]
44
+
@@ -24,24 +24,24 @@ VAR_SAMP(col) -- sample variance
24
24
  -- Exact (sorts all data — can be slow on large sets)
25
25
  PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY col) -- median
26
26
  PERCENTILE_DISC(0.25) WITHIN GROUP (ORDER BY col) -- 25th percentile (discrete)
27
-
28
- -- Approximate (fast, uses HLL sketch — Vantage)
29
- APPROX_PERCENTILE(col, 0.5)
30
- APPROX_PERCENTILE(col, 0.95)
27
+ -- No native approximate percentile function found on Teradata 20.00.30.76.
31
28
  ```
32
29
 
33
- ## Approximate Count Distinct
30
+ ## Distinct Count
34
31
  ```sql
35
- -- HyperLogLog-based — much faster than COUNT(DISTINCT) on large data
36
- APPROX_COUNT_DISTINCT(col)
32
+ COUNT(DISTINCT col)
33
+ -- No native APPROX_COUNT_DISTINCT found on Teradata 20.00.30.76.
37
34
  ```
38
35
 
39
36
  ## String Aggregation
40
37
  ```sql
41
- -- Concatenate values into one string (XML-based, common pattern)
42
- XMLAGG(XMLELEMENT(NAME x, col || ',') ORDER BY col)
43
-
44
- -- Cleaner alternative using TD_SYSFNLIB.XMLAGG or custom UDF
38
+ -- Concatenate values into one delimited string: aggregate, cast, and trim the trailing comma
39
+ -- Note: wrapping with XMLELEMENT keeps XML tags in output — use TRIM + CAST instead
40
+ SELECT dept,
41
+ TRIM(TRAILING ',' FROM CAST(XMLAGG(TRIM(col) || ',' ORDER BY col) AS VARCHAR(1000))) AS vals
42
+ FROM db.t
43
+ GROUP BY dept;
44
+ -- LISTAGG and STRING_AGG are not available on Teradata 20.00.30.76.
45
45
  ```
46
46
 
47
47
  ## GROUP BY Variants
@@ -52,9 +52,11 @@ SELECT dept, SUM(salary) FROM db.t GROUP BY dept;
52
52
  -- Multiple grouping sets in one pass
53
53
  GROUP BY GROUPING SETS ((dept), (region), (dept, region), ())
54
54
 
55
- -- Equivalent shorthand
55
+ -- CUBE produces the same result as the four-set GROUPING SETS example above
56
+ GROUP BY CUBE(dept, region) -- all combinations + grand total
57
+
58
+ -- ROLLUP is smaller: (dept, region), (dept), and the grand total () only
56
59
  GROUP BY ROLLUP(dept, region) -- all prefixes + grand total
57
- GROUP BY CUBE(dept, region) -- all combinations
58
60
 
59
61
  -- Identify which grouping a row belongs to
60
62
  GROUPING(dept) -- 1 if dept is aggregated away, 0 if not
@@ -77,7 +79,7 @@ SUM(CASE WHEN status = 'active' THEN 1 ELSE 0 END) AS active_count
77
79
  AVG(NULLIFZERO(amount))
78
80
 
79
81
  -- Max of a filtered subset
80
- MAX(CASE WHEN category = 'A' THEN value END)
82
+ MAX(CASE WHEN category = 'A' THEN "value" END) -- value is a reserved word; must be quoted
81
83
  ```
82
84
 
83
85
  ## Common Patterns
@@ -12,7 +12,6 @@ Discovers frequent patterns (itemsets) and association rules in transactional da
12
12
  ```sql
13
13
  SELECT * FROM TD_Apriori(
14
14
  ON { db.table | db.view | (query) } AS InputTable [ PARTITION BY ANY ]
15
- [ OUT TABLE OutputTable('db.pattern_table') ] -- optional; writes output to a named table
16
15
  USING
17
16
  TargetColumn('item_col') -- required; column containing item data; max 1
18
17
  [ IDColumn('transaction_id_col') ] -- required for sparse input; groups items into transactions
@@ -24,10 +23,20 @@ SELECT * FROM TD_Apriori(
24
23
  [ PatternsOrRules('Patterns') ] -- optional; default 'Patterns'
25
24
  -- 'Patterns' — frequent itemsets only
26
25
  -- 'Rules' — association rules with metrics
27
- -- 'Both' — patterns and rules
26
+ -- 'Both' is rejected (error 7810); run the function twice to get both outputs
28
27
  [ Support(0.01) ] -- optional; min itemset frequency as proportion
29
28
  -- of total transactions; range (0,1]; default 0.01
29
+ -- threshold is inclusive (itemsets at exactly the threshold are returned)
30
30
  ) AS t;
31
+
32
+ -- TD_Apriori has no OUT TABLE clause (error 7810). To save the output, wrap the call in CREATE TABLE AS.
33
+ -- Use MULTISET so repeated rows are not dropped:
34
+ CREATE MULTISET TABLE db.pattern_table AS (
35
+ SELECT * FROM TD_Apriori(
36
+ ON db.baskets AS InputTable PARTITION BY ANY
37
+ USING TargetColumn('item') IDColumn('txn') Support(0.2)
38
+ ) AS t
39
+ ) WITH DATA;
31
40
  ```
32
41
 
33
42
  **Input formats:**
@@ -37,7 +46,22 @@ SELECT * FROM TD_Apriori(
37
46
  | Dense | One row per transaction; items in a single delimited cell in `TargetColumn` | Not used |
38
47
  | Sparse | One row per item; `IDColumn` groups items into a transaction | Required |
39
48
 
40
- **Output — Patterns** (`PatternsOrRules('Patterns')` or `'Both'`):
49
+ > **Character set:** `TargetColumn` must be a LATIN column. Unicode input causes `Itemset` values to contain NUL characters between every letter (e.g. `CHAR_LENGTH('bread,milk')` returns 19 instead of 10). Convert Unicode columns first:
50
+ > ```sql
51
+ > SELECT * FROM TD_Apriori(
52
+ > ON (SELECT TRANSLATE(items USING UNICODE_TO_LATIN) AS items FROM db.baskets) AS InputTable PARTITION BY ANY
53
+ > USING TargetColumn('items') InputDelimiter(',') Support(0.2)
54
+ > ) AS t;
55
+ > ```
56
+ > This was observed on dense input. Sparse Unicode input was not tested.
57
+
58
+ > **Single-item itemsets:** Without `PartitionColumns`, `Patterns` output contains only itemsets of 2 to `MaxLen` items — single-item itemsets are not returned. With `PartitionColumns`, single items above the support threshold are included. To compute single-item support without a partition, use a separate query:
59
+ > ```sql
60
+ > SELECT item, COUNT(DISTINCT txn) * 1.0 / (SELECT COUNT(DISTINCT txn) FROM db.baskets) AS support
61
+ > FROM db.baskets GROUP BY item;
62
+ > ```
63
+
64
+ **Output — Patterns** (`PatternsOrRules('Patterns')`):
41
65
 
42
66
  | Column | Type | Description |
43
67
  |--------|------|-------------|
@@ -45,15 +69,15 @@ SELECT * FROM TD_Apriori(
45
69
  | `Itemset` | VARCHAR | Comma-separated (or `OutputDelimiter`) set of items |
46
70
  | `Support` | REAL | Proportion of transactions containing this itemset |
47
71
 
48
- **Output — Rules** (`PatternsOrRules('Rules')` or `'Both'`):
72
+ **Output — Rules** (`PatternsOrRules('Rules')`):
49
73
 
50
74
  | Column | Type | Description |
51
75
  |--------|------|-------------|
52
76
  | `partition_column(s)` | ANY | Partition pass-through columns |
53
77
  | `Antecedent` | VARCHAR | The "if" item(s) in the rule |
54
78
  | `Consequence` | VARCHAR | The "then" item(s) in the rule |
55
- | `Ante_ItemCnt` | — | Item count for antecedent (undocumented in source) |
56
- | `Conse_Item_Cnt` | — | Item count for consequence (undocumented in source) |
79
+ | `Ante_Item_Cnt` | INTEGER | Number of items in the antecedent |
80
+ | `Conse_Item_Cnt` | INTEGER | Number of items in the consequence |
57
81
  | `cntb` | INTEGER | Co-occurrence count of antecedent and consequence |
58
82
  | `Cnt_Antecedent` | INTEGER | Occurrence count of antecedent |
59
83
  | `Cnt_Consequence` | INTEGER | Occurrence count of consequence |
@@ -82,7 +106,8 @@ SELECT * FROM TD_CFilter(
82
106
  [ PartitionColumns('region_col') ] -- optional; partition input and output; max 10
83
107
  -- output is nondeterministic unless each
84
108
  -- partition_col is unique within TransactionIDColumns
85
- [ MaxDistinctItems(100) ] -- optional; max item set size; default 100
109
+ [ MaxDistinctItems(100) ] -- optional; maximum number of distinct items in one transaction; default 100
110
+ -- transactions with more items are silently excluded from all counts
86
111
  ) AS t;
87
112
  ```
88
113
 
@@ -100,9 +125,10 @@ SELECT * FROM TD_CFilter(
100
125
  | `support` | REAL | `cntb / tran_cnt` — proportion of transactions containing both items |
101
126
  | `confidence` | REAL | `cntb / cnt1` — proportion of item1 transactions that also contain item2 |
102
127
  | `lift` | REAL | `support(A∩B) / (support(A) × support(B))` — >1 positive, =1 independent, <1 negative |
103
- | `z_score` | REAL | `(cntb - mean(cntb)) / sd(cntb)` — co-occurrence significance; not calculated if all `cntb` values are equal |
128
+ | `z_score` | REAL | `(cntb - mean(cntb)) / population_sd(cntb)` — co-occurrence significance; not calculated if all `cntb` values are equal |
104
129
 
105
130
  > **Notes:**
106
131
  > - Always pairwise — unlike `TD_Apriori`, does not produce itemsets of size > 2 or association rules
107
132
  > - `TransactionIDColumns` accepts multiple columns (composite transaction key)
108
- > - `z_score` is null/absent when all `cntb` values are equal (sd = 0)
133
+ > - **Each pair appears in both directions** — `(bread, milk)` and `(milk, bread)` are separate rows, with `cnt1`, `cnt2`, and `confidence` swapped; the guide's "one row per item pair" is inaccurate
134
+ > - `z_score` uses the **population** standard deviation (not sample); it is null/absent when all `cntb` values are equal (sd = 0)
@@ -0,0 +1,259 @@
1
+ # Teradata Authorization Objects
2
+
3
+ Authorization objects store external service credentials securely inside Teradata, decoupling
4
+ secrets from SQL code. They are used by NOS table operators (`READ_NOS`, `WRITE_NOS`),
5
+ external object access via `CREATE FOREIGN TABLE` and `CREATE DATALAKE`, and AI inference
6
+ functions (`AI_TextEmbeddings` and others).
7
+
8
+ Rather than embedding credentials directly in SQL, you reference a named authorization object.
9
+ Credentials are stored once, managed centrally, and access is controlled via standard
10
+ Teradata permissions.
11
+
12
+ ---
13
+
14
+ ## CREATE / REPLACE AUTHORIZATION
15
+
16
+ ```sql
17
+ { CREATE | REPLACE } AUTHORIZATION [DatabaseName.]authorization_name
18
+ [ AS { DEFINER | INVOKER } TRUSTED ]
19
+ { user_password_auth | extended_auth }
20
+
21
+ -- Form 1: user_password_auth
22
+ USER 'user_value'
23
+ PASSWORD 'password_value'
24
+ [ SESSION_TOKEN 'session_token_value' ]
25
+
26
+ -- Form 2: extended_auth (AWS IAM role assumption only)
27
+ USING AUTHSERVICETYPE 'ASSUME_ROLE'
28
+ ROLENAME 'arn:aws:iam::account-id:role/role-name'
29
+ EXTERNALID 'external_id_value'
30
+ [ DURATION_SECONDS 'duration_in_seconds' ]
31
+ ```
32
+
33
+ **`AS DEFINER` / `AS INVOKER` / `TRUSTED`:**
34
+
35
+ | Clause | Meaning |
36
+ |--------|---------|
37
+ | *(none — plain)* | **Recommended for DATALAKE and foreign tables.** Can be created in any database the creator has `CREATE AUTHORIZATION` on (including global databases). Referenced as `EXTERNAL SECURITY <db>.<auth>` with no keywords. |
38
+ | `AS DEFINER` | Shared access. Auth must be in the **same database** as the referencing object. Referenced as `EXTERNAL SECURITY DEFINER TRUSTED <auth>` (unqualified only — Error 3706 if qualified). |
39
+ | `AS INVOKER` | Exclusive access by the creating user; created in the user's own database. Referenced as `EXTERNAL SECURITY INVOKER TRUSTED <auth>`. |
40
+ | `TRUSTED` | Required **only** when using the `AS DEFINER` or `AS INVOKER` forms in an `EXTERNAL SECURITY` clause. Not needed for plain auth. |
41
+
42
+ **Form 2 — ASSUME_ROLE:**
43
+ - `AUTHSERVICETYPE` — only supported value is `'ASSUME_ROLE'`
44
+ - `ROLENAME` — ARN of the IAM role to assume
45
+ - `EXTERNALID` — external ID required by the IAM role trust policy
46
+ - `DURATION_SECONDS` — session duration 900–43200 seconds; default 3600; must not exceed the IAM role's max session duration
47
+
48
+ > **EXTERNAL SECURITY — the reference must match how the auth object was created:**
49
+ >
50
+ > | Auth created as | `EXTERNAL SECURITY` reference | Qualified `db.auth` allowed? |
51
+ > |---|---|---|
52
+ > | Plain (no `AS` clause) | `EXTERNAL SECURITY <db>.<auth>` (no keywords) | **Yes** — verified for DATALAKE and FOREIGN TABLE |
53
+ > | `AS DEFINER TRUSTED` | `EXTERNAL SECURITY DEFINER TRUSTED <auth>` | **No** — Error 3706; auth must be in the same DB as the object |
54
+ > | `AS INVOKER TRUSTED` | `EXTERNAL SECURITY INVOKER TRUSTED <auth>` | No |
55
+ >
56
+ > A mismatch raises Error 6953 (`authorization definition does not match`) or Error 3706.
57
+ >
58
+ > **Elastic Compute:** For DATALAKE objects, the auth object must be in a GLOBAL database
59
+ > so it replicates across all CE instances. For NOS foreign tables, place it in the same
60
+ > local or global database as the table.
61
+ >
62
+ > **GRANT EXECUTE:** After creating an auth object, grant `EXECUTE` so roles and databases
63
+ > can use it — see [Permissions](#permissions) below.
64
+
65
+ ---
66
+
67
+ ## NOS and OTF External Security Authorization
68
+
69
+ Used with `CREATE FOREIGN TABLE`, `CREATE DATALAKE`, `READ_NOS`, and `WRITE_NOS`.
70
+ The `USER` and `PASSWORD` fields map to object storage credentials, not API keys.
71
+
72
+ ### Field Mapping — Object Storage Providers
73
+
74
+ | Storage system | USER | PASSWORD | SESSION_TOKEN |
75
+ |---|---|---|---|
76
+ | AWS (IAM user) | Access Key ID | Access Key Secret | Session Token *(optional — STS temporary creds only)* |
77
+ | Azure Shared Key | Storage Account Name | Storage Account Key | — |
78
+ | Azure SAS | Storage Account Name | Account SAS Token | — |
79
+ | GCS (S3 interop mode) | Access Key ID | Access Key Secret | — |
80
+ | GCS (native) | Client Email | Private Key | — |
81
+ | On-premises object store | Access Key ID | Access Key Secret | — |
82
+ | Public bucket | `''` (empty string) | `''` (empty string) | — |
83
+
84
+ ### Examples
85
+
86
+ ```sql
87
+ -- AWS: plain auth for OTF DATALAKE (no AS clause — plain auth + qualified ref is verified)
88
+ CREATE AUTHORIZATION global_db.aws_glue_auth
89
+ USER '<aws_access_key_id>'
90
+ PASSWORD '<aws_access_key_secret>';
91
+
92
+ -- AWS: ASSUME_ROLE for OTF DATALAKE or NOS
93
+ CREATE AUTHORIZATION global_db.aws_role_auth
94
+ USING AUTHSERVICETYPE 'ASSUME_ROLE'
95
+ ROLENAME '<arn:aws:iam::account-id:role/role-name>'
96
+ EXTERNALID '<external_id>'
97
+ DURATION_SECONDS '3600';
98
+
99
+ -- Azure Shared Key (Storage Account Name + Storage Account Key)
100
+ CREATE AUTHORIZATION global_db.azure_auth
101
+ USER '<storage_account_name>'
102
+ PASSWORD '<storage_account_key>';
103
+
104
+ -- Azure SAS (Storage Account Name + Account SAS Token)
105
+ CREATE AUTHORIZATION global_db.azure_sas_auth
106
+ USER '<storage_account_name>'
107
+ PASSWORD '<account_sas_token>';
108
+
109
+ -- GCS (S3 interop mode — Access Key ID + Secret)
110
+ CREATE AUTHORIZATION global_db.gcs_interop_auth
111
+ USER '<gcs_access_key_id>'
112
+ PASSWORD '<gcs_access_key_secret>';
113
+
114
+ -- GCS (native — Client Email + Private Key)
115
+ CREATE AUTHORIZATION global_db.gcs_native_auth
116
+ USER '<client_email>'
117
+ PASSWORD '<private_key>';
118
+
119
+ -- Public bucket (empty credentials)
120
+ CREATE AUTHORIZATION global_db.public_auth
121
+ USER ''
122
+ PASSWORD '';
123
+
124
+ -- NOS FOREIGN TABLE: AS DEFINER TRUSTED, auth in same DB as table, unqualified reference
125
+ CREATE AUTHORIZATION mydb.s3_ft_auth
126
+ AS DEFINER TRUSTED
127
+ USER '<aws_access_key_id>'
128
+ PASSWORD '<aws_access_key_secret>';
129
+
130
+ CREATE MULTISET FOREIGN TABLE mydb.orders_ft,
131
+ EXTERNAL SECURITY DEFINER TRUSTED s3_ft_auth -- unqualified; qualified raises Error 3706
132
+ USING ( LOCATION ('/S3/s3.amazonaws.com/<bucket>/<prefix>/') STOREDAS ('PARQUET') )
133
+ NO PRIMARY INDEX;
134
+
135
+ -- OTF DATALAKE: plain auth in GLOBAL database, qualified reference, no keywords
136
+ CREATE DATALAKE my_lake
137
+ EXTERNAL SECURITY CATALOG global_db.aws_glue_auth,
138
+ EXTERNAL SECURITY STORAGE global_db.aws_glue_auth
139
+ USING catalog_type ('glue') ... TABLE FORMAT iceberg;
140
+ ```
141
+
142
+ ---
143
+
144
+ ## AI Function Authorization
145
+
146
+ Used with `AI_TextEmbeddings`, `AI_Generate`, and other AI inference functions. The `USER`,
147
+ `PASSWORD`, and `SESSION_TOKEN` fields map to AI provider API credentials — **different
148
+ semantics from object storage auth above.**
149
+
150
+ ### Field Mapping — AI Inference Providers
151
+
152
+ | Provider | USER | PASSWORD | SESSION_TOKEN |
153
+ |----------|------|----------|---------------|
154
+ | **AWS Bedrock** | Access Key ID | Access Key Secret | Session Token *(optional — STS only)* |
155
+ | **Azure OpenAI** | API Base URL (endpoint) | API Key | API Version *(required)* |
156
+ | **Google Cloud (Vertex)** | Project | Region | Access Token *(required)* |
157
+ | **NVIDIA NIM** | API Base URL | API Key | *(not used)* |
158
+ | **LiteLLM** | API Base URL | API Key | *(not used)* |
159
+
160
+ ### Examples
161
+
162
+ ```sql
163
+ -- AWS Bedrock (long-term IAM user — omit SESSION_TOKEN)
164
+ CREATE AUTHORIZATION db.td_gen_aws_auth
165
+ USER '<aws_access_key_id>'
166
+ PASSWORD '<aws_access_key_secret>';
167
+
168
+ -- AWS Bedrock (STS temporary credentials — include SESSION_TOKEN)
169
+ CREATE AUTHORIZATION db.td_gen_aws_auth
170
+ USER '<aws_access_key_id>'
171
+ PASSWORD '<aws_access_key_secret>'
172
+ SESSION_TOKEN '<aws_session_token>';
173
+
174
+ -- AWS Bedrock via IAM role assumption
175
+ CREATE AUTHORIZATION db.td_gen_aws_role_auth
176
+ USING AUTHSERVICETYPE 'ASSUME_ROLE'
177
+ ROLENAME 'arn:aws:iam::123456789012:role/teradata-bedrock-role'
178
+ EXTERNALID '<external_id>'
179
+ DURATION_SECONDS '3600';
180
+
181
+ -- Azure OpenAI (SESSION_TOKEN = API version string — required)
182
+ CREATE AUTHORIZATION db.td_gen_azure_auth
183
+ USER '<api_base_url>'
184
+ PASSWORD '<api_key>'
185
+ SESSION_TOKEN '2024-02-15-preview';
186
+
187
+ -- Google Cloud Vertex AI (SESSION_TOKEN = Access Token — required)
188
+ CREATE AUTHORIZATION db.td_gen_gcp_auth
189
+ USER '<gcp_project>'
190
+ PASSWORD '<gcp_region>'
191
+ SESSION_TOKEN '<gcp_access_token>';
192
+
193
+ -- NVIDIA NIM
194
+ CREATE AUTHORIZATION db.td_gen_nim_auth
195
+ USER '<nim_api_base_url>'
196
+ PASSWORD '<nim_api_key>';
197
+
198
+ -- LiteLLM
199
+ CREATE AUTHORIZATION db.td_gen_litellm_auth
200
+ USER '<litellm_base_url>'
201
+ PASSWORD '<litellm_api_key>';
202
+
203
+ -- Rotate credentials without dropping (use REPLACE)
204
+ REPLACE AUTHORIZATION db.td_gen_aws_auth
205
+ USER '<new_access_key>'
206
+ PASSWORD '<new_secret_key>';
207
+ ```
208
+
209
+ ---
210
+
211
+ ## Permissions
212
+
213
+ ### EXECUTE — allow a role or database to use an auth object
214
+
215
+ ```sql
216
+ -- Grant to a role (preferred — access follows role membership)
217
+ GRANT EXECUTE ON db.my_auth TO analyst_role;
218
+ GRANT EXECUTE ON db.my_auth TO analyst_role WITH GRANT OPTION;
219
+
220
+ -- Grant at database level (covers all auth objects in the DB)
221
+ GRANT EXECUTE ON db TO analyst_role;
222
+
223
+ -- Revoke
224
+ REVOKE EXECUTE ON db.my_auth FROM analyst_role;
225
+ ```
226
+
227
+ ### CREATE / DROP AUTHORIZATION
228
+
229
+ ```sql
230
+ GRANT CREATE AUTHORIZATION ON db TO dba_user;
231
+ GRANT DROP AUTHORIZATION ON db TO dba_user;
232
+ REVOKE CREATE AUTHORIZATION ON db FROM dba_user;
233
+ ```
234
+
235
+ ### Permissions Inheritance — Critical for Operational Pipelines
236
+
237
+ If a **view or macro in database B** references an auth object in **database A**, then
238
+ **database B must have `EXECUTE`** — not just the user running the query.
239
+
240
+ ```sql
241
+ -- Works in dev (user has EXECUTE) but the operational view will fail without this:
242
+ GRANT EXECUTE ON credentials_db.my_auth TO operational_db;
243
+ -- ^^^^^^^^^^^^^^
244
+ -- Grant to the DATABASE, not just the user
245
+ ```
246
+
247
+ **Rule of thumb:** store auth objects in a dedicated credentials database, grant `EXECUTE`
248
+ to every database that references them, and grant to roles rather than individual users.
249
+
250
+ ---
251
+
252
+ ## DROP AUTHORIZATION
253
+
254
+ ```sql
255
+ DROP AUTHORIZATION [DatabaseName.]authorization_name;
256
+ ```
257
+
258
+ > Dropping an auth object immediately breaks any query, view, or procedure that references it.
259
+ > Prefer `REPLACE AUTHORIZATION` for credential rotation.
@@ -37,16 +37,25 @@ Teradata uses a **big-endian** model. The most significant byte is on the left;
37
37
 
38
38
  ## Mismatched-Length Operands (BITAND / BITOR / BITXOR only)
39
39
 
40
- When `target_arg` and `bit_mask_arg` have different lengths, they are **aligned on the LSB** and the shorter argument is **left-padded with zeros**:
40
+ The padding direction depends on the operand types:
41
+
42
+ - **Numeric target + shorter hex mask:** the mask is **right-padded** with zeros, so it acts on the **leading (upper) bytes**.
43
+ - **Two BYTE/VARBYTE operands of different lengths:** aligned on the **LSB**; the shorter is **left-padded**.
44
+
45
+ Always write hex masks at full width to avoid ambiguity:
41
46
 
42
47
  ```sql
43
- -- 'FFFF'XB (2 bytes) vs INTEGER (4 bytes)
44
- -- Effective operation: '0000FFFF'XB AND '11223344'XB
45
- SELECT BITAND(287454020, 'FFFF'XB); -- masks lower 2 bytes only
46
- -- Result: 0x3344 = 13124
48
+ -- Short mask is RIGHT-padded: 'FFFF'XB acts as 'FFFF0000'XB → upper two bytes
49
+ SELECT BITAND(287454020, 'FFFF'XB); -- 287440896 = 0x11220000 (the UPPER two bytes)
50
+
51
+ -- Full-width mask: targets the lower two bytes as intended
52
+ SELECT BITAND(287454020, '0000FFFF'XB); -- 13124 = 0x3344
53
+
54
+ -- Two BYTE/VARBYTE operands: aligned on LSB (shorter is left-padded)
55
+ SELECT BITAND('0F0F'XB, 'FF'XB); -- '000F'XB
47
56
  ```
48
57
 
49
- > Left-padding for operands vs right-padding for literals — they go in opposite directions. Always visualize the alignment on the **least significant byte**.
58
+ > This applies to BITAND, BITOR, and BITXOR. Hex literals are always right-padded when extended to a required size (see above) — the same direction as the numeric+mask case.
50
59
 
51
60
  ---
52
61
 
@@ -104,9 +113,9 @@ SELECT BITAND(flags_col, BITNOT(CAST(8 AS INTEGER))) AS flags_without_bit3 FROM
104
113
  -- Toggle a bit (XOR with the bit mask)
105
114
  SELECT BITXOR(flags_col, 8) AS toggled FROM db.t;
106
115
 
107
- -- Mask lower 2 bytes of an INTEGER using a hex literal
108
- SELECT BITAND(some_int_col, 'FFFF'XB) AS lower_two_bytes FROM db.t;
109
- -- 'FFFF'XB is left-padded to '0000FFFF'XB to match INTEGER width
116
+ -- Mask lower 2 bytes of an INTEGER — use a full-width mask; short mask is right-padded
117
+ SELECT BITAND(some_int_col, '0000FFFF'XB) AS lower_two_bytes FROM db.t;
118
+ -- 'FFFF'XB alone would act as '0000FFFF'XB is NOT correct: it acts as 'FFFF0000'XB (upper bytes)
110
119
  ```
111
120
 
112
121
  ---
@@ -156,10 +165,10 @@ SELECT SUBBITSTR(20, 2, 3);
156
165
  -- Bits at positions 2,3,4 = 101 → right-justified in VARBYTE(1) = 00000101 = 5
157
166
 
158
167
  -- Extract 8 bits starting at bit 4 from an INTEGER
159
- SELECT SUBBITSTR(some_int_col, 4, 8) FROM db.t; -- returns VARBYTE(1)
168
+ SELECT SUBBITSTR(some_int_col, 4, 8) FROM db.t; -- value is 1 byte; declared type is VARBYTE(4) for INTEGER input
160
169
 
161
- -- Extract 12 bits — rounded up to 2-byte VARBYTE
162
- SELECT SUBBITSTR(some_int_col, 8, 12) FROM db.t; -- returns VARBYTE(2), right-justified
170
+ -- Extract 12 bits — value is 2 bytes, right-justified
171
+ SELECT SUBBITSTR(some_int_col, 8, 12) FROM db.t; -- declared VARBYTE(4); value is 2 bytes right-justified
163
172
  ```
164
173
 
165
174
  > **Output size:** always rounded up to the next byte boundary. 3 bits requested → 1-byte VARBYTE. 9 bits requested → 2-byte VARBYTE. Excess bits are zero-filled on the left.
@@ -217,8 +226,8 @@ SELECT ROTATERIGHT(4, 4); -- 00000100 rotated right 4 = 01000000 = 64 (wrapp
217
226
  SELECT TO_BYTE(23); -- BYTEINT 23 → BYTE(1) = '17'XB
218
227
  SELECT TO_BYTE(287454020); -- INTEGER → BYTE(4) = '11223344'XB
219
228
 
220
- -- Useful for inspecting the hex value of a numeric result
221
- SELECT TO_BYTE(CAST(BITAND(some_col, 'FFFF'XB) AS INTEGER)) FROM db.t;
229
+ -- Useful for inspecting the hex value of a numeric result; use a full-width mask
230
+ SELECT TO_BYTE(CAST(BITAND(some_col, '0000FFFF'XB) AS INTEGER)) FROM db.t;
222
231
  ```
223
232
 
224
233
  | Input type | Output type |
@@ -234,7 +243,8 @@ SELECT TO_BYTE(CAST(BITAND(some_col, 'FFFF'XB) AS INTEGER)) FROM db.t;
234
243
 
235
244
  - **All functions are Teradata-specific** — there are no ANSI SQL equivalents. Standard SQL has no bitwise function syntax; do not use `&`, `|`, `^`, `~` operators.
236
245
  - **Signed integer overflow:** shifting or rotating a bit into the MSB (Bit 7/15/31/63) makes the result negative — all integers in Vantage are signed. This is expected behavior, not an error.
237
- - **UDF implicit type conversion rules** are stricter than normal Vantage rules. If an argument type doesn't match exactly, use explicit `CAST`. Example: `BITAND(CAST(some_col AS INTEGER), '0000FFFF'XB)`.
246
+ - **UDF implicit type conversion rules** are stricter than normal Vantage rules. A numeric mask may be **narrower** than the target (`BITAND(CAST(255 AS INTEGER), 8)` works) but not **wider** — a wider mask fails with error 9881 (`BITAND(23, CAST(20 AS INTEGER))` where 23 is BYTEINT and the mask is INTEGER). CAST the target to the wider type first: `BITAND(CAST(some_col AS INTEGER), '0000FFFF'XB)`.
238
247
  - **`TD_SYSFNLIB.` prefix** is optional but can be included for clarity: `TD_SYSFNLIB.BITAND(...)`.
239
248
  - **VARBYTE max size** for all functions is 8192 bytes.
240
249
  - **DECIMAL inputs to BITAND** are implicitly converted to `NUMBER(38,0)`. BITNOT does not support DECIMAL/NUMBER.
250
+ - **VARBYTE results cannot be cast to a number.** `CAST(SUBBITSTR(...) AS INTEGER)` fails with error 3532 ("Conversion between BYTE data and other types is illegal"). Compare VARBYTE results as bytes; `BITAND` with a numeric target returns INTEGER, which can be cast.
@@ -12,7 +12,7 @@ ORDER BY DatabaseName;
12
12
  -- DBKind: 'D' = database, 'U' = user
13
13
 
14
14
  -- List users
15
- SELECT UserName, DefaultDatabase, CreateDate
15
+ SELECT UserName, DefaultDatabase, CreateTimeStamp
16
16
  FROM DBC.UsersV
17
17
  ORDER BY UserName;
18
18
  ```
@@ -24,7 +24,7 @@ SELECT TableName, TableKind, CreateTimeStamp, LastAlterTimeStamp
24
24
  FROM DBC.TablesV
25
25
  WHERE DatabaseName = 'mydb'
26
26
  ORDER BY TableKind, TableName;
27
- -- TableKind: 'T'=table, 'V'=view, 'O'=NoPI table, 'Q'=queue, 'E'=error, 'I'=join index
27
+ -- TableKind: 'T'=table, 'V'=view, 'O'=NoPI table, 'Q'=queue, 'E'=external stored procedure, 'I'=join index, 'K'=foreign server object (DATALAKE)
28
28
 
29
29
  -- Search for tables by name pattern
30
30
  SELECT DatabaseName, TableName, TableKind
@@ -66,15 +66,16 @@ SELECT IndexName, IndexType, UniqueFlag, ColumnName, ColumnPosition
66
66
  FROM DBC.IndicesV
67
67
  WHERE DatabaseName = 'mydb' AND TableName = 'mytable'
68
68
  ORDER BY IndexType, ColumnPosition;
69
- -- IndexType: 'P'=primary, 'S'=secondary, 'U'=unique, 'J'=join index
69
+ -- IndexType: 'P'=primary index, 'Q'=partitioned PI, 'S'=secondary, 'U'=unique constraint, 'J'=join index, 'K'=primary key constraint
70
+ -- Uniqueness is in UniqueFlag ('Y'/'N'), not in IndexType
70
71
  ```
71
72
 
72
73
  ## Table Size & Storage
73
74
  ```sql
74
- -- Table sizes in a database (from collected stats — no full scan)
75
+ -- Table sizes in a database (live space accounting — updated on every DML, not from stats)
75
76
  SELECT TableName,
76
- SUM(CurrentPerm) / 1024 / 1024 AS size_mb,
77
- SUM(PeakPerm) / 1024 / 1024 AS peak_mb
77
+ SUM(CurrentPerm) / 1024.0 / 1024.0 AS size_mb,
78
+ SUM(PeakPerm) / 1024.0 / 1024.0 AS peak_mb
78
79
  FROM DBC.TableSizeV
79
80
  WHERE DatabaseName = 'mydb'
80
81
  GROUP BY TableName
@@ -92,18 +93,24 @@ ORDER BY used_gb DESC;
92
93
  ## Statistics
93
94
  ```sql
94
95
  -- Collected statistics on a table
95
- SELECT ColumnName, StatsType, LastCollectTimeStamp, SampleSize
96
+ SELECT ColumnName, StatsType, LastCollectTimeStamp, SampleSizePct
96
97
  FROM DBC.StatsV
97
98
  WHERE DatabaseName = 'mydb' AND TableName = 'mytable'
98
99
  ORDER BY ColumnName;
100
+ -- ColumnName is NULL for multi-column or index stats; StatsType: 'T'=table, 'O'=index
99
101
  ```
100
102
 
101
103
  ## Access Rights
102
104
  ```sql
103
- -- What access does the current user have on a database?
105
+ -- Rights of the current user (DBC.UserRightsV is already scoped to the current user — no UserName column)
104
106
  SELECT AccessRight, DatabaseName, TableName
105
107
  FROM DBC.UserRightsV
106
- WHERE UserName = USER
108
+ ORDER BY DatabaseName, TableName;
109
+
110
+ -- Rights of any user (requires SELECT on DBC)
111
+ SELECT AccessRight, UserName, DatabaseName, TableName
112
+ FROM DBC.AllRightsV
113
+ WHERE UserName = 'someuser'
107
114
  ORDER BY DatabaseName, TableName;
108
115
  ```
109
116
 
@@ -159,8 +166,8 @@ ORDER BY DatalakeName;
159
166
  -- String values include embedded single quotes, e.g. CatalogType returns 'glue' not glue
160
167
  ```
161
168
 
162
- Columns include: `DatalakeName`, `OTFTableFormat` (ICEBERG/DELTA), `CatalogType`
163
- (hive/glue/unity/rest/fabric/biglake), `CatalogLocation` (catalog URL / AWS region),
169
+ Columns include: `DatalakeName`, `OTFTableFormat` (ICEBERG/DELTALAKE), `CatalogType`
170
+ (hive/glue/unity/rest/fabric/biglake), `CatalogLocation` (catalog endpoint URL — not an AWS region string),
164
171
  `StorageLocation`, `StorageEndPoint`, `StorageRegion`, `UnityCatalogName`, `StorageAccountName`.
165
172
 
166
173
  Once a datalake name is known, get its full DDL (auth names, catalog config):
@@ -179,7 +186,7 @@ SELECT ServerName, DataBaseName, AuthorizationName, AuthorizationType,
179
186
  FROM DBC.ServerV
180
187
  ORDER BY ServerName;
181
188
  -- DataBaseName: the database this server is registered in (usually TD_SERVER_DB)
182
- -- TableFormat: ICEBERG, DELTA, or blank for non-OTF servers
189
+ -- TableFormat: ICEBERG, DELTALAKE, or NULL for non-OTF servers
183
190
  -- AuthorizationType: may appear blank on some deployments
184
191
  ```
185
192
 
@@ -188,20 +195,22 @@ ORDER BY ServerName;
188
195
  ### DBC.ManagedOTFTablesV — Managed OTF Tables
189
196
 
190
197
  ```sql
191
- SELECT DatalakeName, DatabaseName, TableName, TableFormat,
192
- CreateTimeStamp, LastAlterTimeStamp
198
+ SELECT DatalakeName, DatabaseName, TableName, OTFTableFormat,
199
+ CatalogDBName, CatalogTableName
193
200
  FROM DBC.ManagedOTFTablesV
194
201
  ORDER BY DatalakeName, DatabaseName, TableName;
202
+ -- No timestamp columns; OTFTableFormat values: ICEBERG, DELTALAKE
195
203
  ```
196
204
 
197
205
  ### DBC.OtfStatsV — OTF Statistics
198
206
 
199
207
  ```sql
200
- SELECT DatabaseName, TableName, ColumnName, StatsType,
201
- LastCollectTimeStamp, SampleSize
208
+ SELECT CatalogName, TableName, ColumnName, StatsType,
209
+ LastCollectTimeStamp, SampleSizePct
202
210
  FROM DBC.OtfStatsV
203
- WHERE DatabaseName = 'my_lake'
211
+ WHERE CatalogName = 'my_lake'
204
212
  ORDER BY TableName, ColumnName;
213
+ -- Filter on CatalogName (not DatabaseName); SampleSizePct not SampleSize
205
214
  ```
206
215
 
207
216
  ### DBC.AllStatsV — All Statistics (Teradata + OTF)