riftdiff 0.1.0__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.
riftdiff-0.1.0/LICENSE ADDED
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Chetan Chirag
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -0,0 +1,301 @@
1
+ Metadata-Version: 2.4
2
+ Name: riftdiff
3
+ Version: 0.1.0
4
+ Summary: Local-first data diff, powered by DuckDB. See exactly what changed before you merge.
5
+ Author: Chetan Chirag
6
+ License-Expression: MIT
7
+ Project-URL: Homepage, https://github.com/ChetanChirag24/rift
8
+ Project-URL: Source, https://github.com/ChetanChirag24/rift
9
+ Project-URL: Issues, https://github.com/ChetanChirag24/rift/issues
10
+ Project-URL: Changelog, https://github.com/ChetanChirag24/rift/releases
11
+ Keywords: diff,data,csv,parquet,duckdb,data-quality,git
12
+ Classifier: Development Status :: 4 - Beta
13
+ Classifier: Environment :: Console
14
+ Classifier: Intended Audience :: Developers
15
+ Classifier: Operating System :: OS Independent
16
+ Classifier: Programming Language :: Python :: 3
17
+ Classifier: Programming Language :: Python :: 3 :: Only
18
+ Classifier: Programming Language :: Python :: 3.10
19
+ Classifier: Programming Language :: Python :: 3.11
20
+ Classifier: Programming Language :: Python :: 3.12
21
+ Classifier: Programming Language :: Python :: 3.13
22
+ Classifier: Programming Language :: Python :: 3.14
23
+ Classifier: Topic :: Database
24
+ Classifier: Topic :: Software Development :: Quality Assurance
25
+ Requires-Python: >=3.10
26
+ Description-Content-Type: text/markdown
27
+ License-File: LICENSE
28
+ Requires-Dist: duckdb>=0.10.0
29
+ Requires-Dist: rich>=13.0.0
30
+ Provides-Extra: dev
31
+ Requires-Dist: pytest>=7; extra == "dev"
32
+ Dynamic: license-file
33
+
34
+ <p align="center">
35
+ <img src="https://raw.githubusercontent.com/ChetanChirag24/rift/719315b1c043077b447e6c6ff89f3b1db84befa5/docs/assets/hero.svg" alt="rift: two stacked data layers, before and after, with a scanner finding added, removed and changed rows" width="100%">
36
+ </p>
37
+
38
+ <p align="center">
39
+ <a href="https://github.com/ChetanChirag24/rift/actions/workflows/ci.yml"><img alt="CI" src="https://github.com/ChetanChirag24/rift/actions/workflows/ci.yml/badge.svg"></a>
40
+ <img alt="Python 3.10+" src="https://img.shields.io/badge/python-3.10%2B-3776AB?logo=python&logoColor=white">
41
+ <img alt="Powered by DuckDB" src="https://img.shields.io/badge/powered%20by-DuckDB-FFF000?logo=duckdb&logoColor=black">
42
+ <img alt="Formats" src="https://img.shields.io/badge/formats-CSV%20%C2%B7%20Parquet%20%C2%B7%20JSON-22d3ee">
43
+ <img alt="Git aware" src="https://img.shields.io/badge/git-aware-F05032?logo=git&logoColor=white">
44
+ <img alt="MIT License" src="https://img.shields.io/badge/license-MIT-a78bfa">
45
+ </p>
46
+
47
+ <p align="center"><b>Think <code>git diff</code>, but for tables.</b><br>
48
+ Point rift at two versions of a dataset and it tells you which rows were <b>added</b>, which were <b>removed</b>, and exactly which values <b>changed</b>, all on your own machine.</p>
49
+
50
+ ---
51
+
52
+ ## Why rift?
53
+
54
+ Code changes get reviewed line by line. **Data changes usually don't.** A CSV export is regenerated, a SQL model is refactored, a price list is updated, and the only check is "the row count looks about right."
55
+
56
+ rift shows you what actually changed:
57
+
58
+ - πŸ” **Row-level answers.** Not just "3 rows differ": you see *which* rows and *which* columns, with before β†’ after values.
59
+ - 🏠 **Local-first.** Your data never leaves your laptop. No server, no account, no upload.
60
+ - ⚑ **Fast.** Both files are loaded into an in-memory [DuckDB](https://duckdb.org) database and compared with SQL, so millions of rows are fine.
61
+ - 🌿 **Git-aware.** Compare a file with any earlier commit: `rift HEAD~1:data.csv data.csv -k id`.
62
+ - πŸ€– **CI-ready.** Exits like `diff`: `0` = identical, `1` = different, `2` = error. Add `--json` for machine-readable output.
63
+
64
+ ## See it in action
65
+
66
+ <p align="center">
67
+ <img src="https://raw.githubusercontent.com/ChetanChirag24/rift/719315b1c043077b447e6c6ff89f3b1db84befa5/docs/assets/terminal.svg" alt="Terminal: rift compares data/customers.csv with its previous git commit and prints added, removed and changed rows" width="100%">
68
+ </p>
69
+
70
+ ## How it works
71
+
72
+ <p align="center">
73
+ <img src="https://raw.githubusercontent.com/ChetanChirag24/rift/719315b1c043077b447e6c6ff89f3b1db84befa5/docs/assets/how-it-works.svg" alt="Diagram: two files (one read from git) flow into a DuckDB cube which joins them on the key, then out to a rich table, JSON and an exit code" width="100%">
74
+ </p>
75
+
76
+ 1. **Load.** rift reads each side into DuckDB. Either side can be a CSV, Parquet or JSON file, or a version of a file from git history (`REV:path`).
77
+ 2. **Check the key.** The key column(s) you pass must exist on both sides and be unique and non-empty, so every row has exactly one partner.
78
+ 3. **Compare.** rift joins the two tables on the key:
79
+ - keys only in *after* β†’ **added**
80
+ - keys only in *before* β†’ **removed**
81
+ - keys on both sides where any column `IS DISTINCT FROM` its partner β†’ **changed** (so `NULL` vs `NULL` correctly counts as equal)
82
+ 4. **Report.** You get a colour table in the terminal, JSON with `--json`, and an exit code your scripts can act on. Added, removed and retyped columns are reported too.
83
+
84
+ ---
85
+
86
+ ## Real-world examples
87
+
88
+ ### 🧾 "Did last night's CRM export silently drop customers?"
89
+
90
+ Your finance dashboard is fed by a nightly export. Today's revenue number looks odd.
91
+
92
+ ```bash
93
+ rift exports/2026-11-15/customers.csv exports/2026-11-16/customers.csv -k customer_id
94
+ ```
95
+
96
+ ```text
97
+ Removed -1 ← Initech vanished from the export
98
+ Changed ~2 ← Globex: status active β†’ churned, revenue 8500 β†’ 0
99
+ ```
100
+
101
+ Two minutes later you know whether it's a real churn or a broken export.
102
+
103
+ ### πŸ› οΈ "I refactored a SQL model. Is the output exactly the same?"
104
+
105
+ You rewrote a 400-line query to make it faster. The output should not change at all.
106
+
107
+ ```bash
108
+ rift old_model.parquet new_model.parquet -k order_id --ignore _loaded_at
109
+ echo $? # 0 β†’ identical, ship it. 1 β†’ look at the diff first.
110
+ ```
111
+
112
+ `--ignore` skips columns that are *supposed* to differ, like load timestamps.
113
+
114
+ ### πŸ›’ "Which prices am I about to change in the store?"
115
+
116
+ Merchandising sent a new price sheet. Before you upload it, check exactly what it changes:
117
+
118
+ ```bash
119
+ rift prices_live.csv prices_new.csv -k sku --summary
120
+ ```
121
+
122
+ ```text
123
+ Changed values by column
124
+ column rows changed
125
+ price 412
126
+ sale_price 37
127
+ ```
128
+
129
+ 412 price changes when you expected 40? Better to catch that now than after customers do.
130
+
131
+ ### πŸ€– "Why did my model's accuracy drop between training runs?"
132
+
133
+ Diff the feature tables from both runs, keyed on a composite key, and pipe the JSON into `jq`:
134
+
135
+ ```bash
136
+ rift features_run41.parquet features_run42.parquet -k user_id,event_date --json \
137
+ | jq '.column_change_counts'
138
+ ```
139
+
140
+ ```json
141
+ { "avg_session_minutes": 120394, "country": 3 }
142
+ ```
143
+
144
+ One feature changed for 120k users. That's your bug.
145
+
146
+ ### 🌍 Multi-region tables with composite keys
147
+
148
+ The same `id` can exist in several regions? Combine columns into one key:
149
+
150
+ ```bash
151
+ rift before.csv after.csv -k region,id
152
+ ```
153
+
154
+ ---
155
+
156
+ ## Works with Git
157
+
158
+ Data files live in git repos too, and rift treats git history as a first-class source.
159
+
160
+ ### Compare with any commit, branch or tag
161
+
162
+ Use git's own `REV:path` syntax (paths are relative to the repo root, or start with `./` for the current directory):
163
+
164
+ ```bash
165
+ rift HEAD~1:data/customers.csv data/customers.csv -k customer_id # what did my last commit change?
166
+ rift main:data/customers.csv data/customers.csv -k customer_id # what does my branch change vs main?
167
+ rift v1.0:exports/users.parquet v2.0:exports/users.parquet -k id # between two releases
168
+ ```
169
+
170
+ ### Use rift as your `git difftool`
171
+
172
+ ```bash
173
+ git config --global difftool.rift.cmd 'rift "$LOCAL" "$REMOTE" --key id'
174
+ git difftool -y -t rift HEAD~1 -- data/customers.csv
175
+ ```
176
+
177
+ ### Catch data changes in pull requests
178
+
179
+ ```mermaid
180
+ flowchart LR
181
+ A[πŸ‘©β€πŸ’» Pull request<br/>changes data/] --> B[βš™οΈ CI runs<br/>rift main:data.csv data.csv]
182
+ B -->|exit 0| C[βœ… Data unchanged<br/>merge freely]
183
+ B -->|exit 1| D[⚠️ Data changed<br/>diff printed in the log]
184
+ B -->|exit 2| E[❌ Bad key or file<br/>fix the input]
185
+ D --> F[πŸ‘€ Reviewer checks<br/>the row-level diff]
186
+ F --> C
187
+ ```
188
+
189
+ A minimal GitHub Actions job:
190
+
191
+ ```yaml
192
+ name: data-diff
193
+ on: pull_request
194
+
195
+ jobs:
196
+ rift:
197
+ runs-on: ubuntu-latest
198
+ steps:
199
+ - uses: actions/checkout@v7
200
+ with:
201
+ fetch-depth: 0 # rift needs the main branch history
202
+ - uses: actions/setup-python@v7
203
+ with:
204
+ python-version: "3.12"
205
+ - run: pip install riftdiff
206
+ - name: Show what this PR changes in the data
207
+ run: rift origin/main:data/customers.csv data/customers.csv -k customer_id
208
+ continue-on-error: true # report the diff without failing the build
209
+ ```
210
+
211
+ Remove `continue-on-error` if any data change should block the merge until someone signs off.
212
+
213
+ ---
214
+
215
+ ## Install
216
+
217
+ ```bash
218
+ pip install riftdiff
219
+ ```
220
+
221
+ Or from source:
222
+
223
+ ```bash
224
+ git clone https://github.com/ChetanChirag24/rift.git && cd rift
225
+ python3 -m venv .venv && source .venv/bin/activate
226
+ pip install -e '.[dev]'
227
+ pytest
228
+ ```
229
+
230
+ > **macOS + iCloud Desktop/Documents:** iCloud can mark files inside `.venv` as hidden, and Python then skips the `.pth` file that makes the editable install work (`ModuleNotFoundError: No module named 'riftdiff'`). Create the venv as `.venv.nosync` (iCloud ignores `*.nosync`) and symlink it: `python3 -m venv .venv.nosync && ln -s .venv.nosync .venv`.
231
+
232
+ ## Usage
233
+
234
+ ```
235
+ rift BEFORE AFTER --key COLUMN [options]
236
+ ```
237
+
238
+ `BEFORE` and `AFTER` are file paths or git revisions (`HEAD~1:path/to/file.csv`).
239
+
240
+ | Option | What it does |
241
+ | --- | --- |
242
+ | `-k, --key COLUMN` | Column that identifies a row. Repeat it or comma-separate it for a composite key: `-k region,id`. |
243
+ | `-i, --ignore COLUMN` | Leave a column out of the comparison, e.g. `-i updated_at`. Repeatable / comma-separated. |
244
+ | `-n, --limit N` | Show at most N rows per section (default 10; `0` shows all). Counts are always exact. |
245
+ | `--summary` | Print only the counts, schema changes and per-column totals. |
246
+ | `--json` | Print the full diff as JSON for scripts and CI. |
247
+ | `--version` | Print the version. |
248
+
249
+ **Supported formats:** CSV/TSV (`.csv`, `.tsv`, `.txt`), Parquet (`.parquet`, `.pq`) and JSON (`.json`, `.jsonl`, `.ndjson`), optionally compressed with `.gz` or `.zst`. The two sides may be different formats.
250
+
251
+ ### What rift reports
252
+
253
+ | | Meaning |
254
+ | --- | --- |
255
+ | 🟒 **Added** | Keys that exist only in `AFTER` |
256
+ | πŸ”΄ **Removed** | Keys that exist only in `BEFORE` |
257
+ | 🟑 **Changed** | Keys on both sides where at least one value differs, with before β†’ after for each changed column |
258
+ | πŸ“Š **By column** | How many rows changed in each column |
259
+ | 🧬 **Schema** | Columns added, removed, or whose type changed. Retyped columns are compared by their text form, so `10` and `"10"` count as equal. |
260
+
261
+ rift refuses keys that are missing, empty or not unique, because a diff on those would be misleading.
262
+
263
+ ### Exit status
264
+
265
+ | Code | Meaning |
266
+ | --- | --- |
267
+ | `0` | No differences |
268
+ | `1` | Differences found |
269
+ | `2` | Error (missing file, bad key, unreadable data…) |
270
+
271
+ ## Python API
272
+
273
+ ```python
274
+ from riftdiff import diff
275
+
276
+ result = diff("HEAD~1:data/orders.csv", "data/orders.parquet",
277
+ key_column=["region", "order_id"],
278
+ ignore_columns=["updated_at"], limit=100)
279
+
280
+ result.summary() # {'rows_before': ..., 'rows_added': ..., ...}
281
+ result.added # list of row dicts
282
+ result.removed # list of row dicts
283
+ for change in result.changed:
284
+ print(change.key, change.changes) # {'region': 'EU', 'order_id': 7} {'total': (120, 135)}
285
+ result.has_differences # bool
286
+ result.to_dict() # JSON-ready dict
287
+ ```
288
+
289
+ Problems with the input raise `riftdiff.RiftError` with a readable message.
290
+
291
+ ## Try it now
292
+
293
+ The repo includes a small example:
294
+
295
+ ```bash
296
+ rift examples/customers_before.csv examples/customers_after.csv -k customer_id
297
+ ```
298
+
299
+ ## License
300
+
301
+ [MIT](https://github.com/ChetanChirag24/rift/blob/719315b1c043077b447e6c6ff89f3b1db84befa5/LICENSE)
@@ -0,0 +1,268 @@
1
+ <p align="center">
2
+ <img src="https://raw.githubusercontent.com/ChetanChirag24/rift/719315b1c043077b447e6c6ff89f3b1db84befa5/docs/assets/hero.svg" alt="rift: two stacked data layers, before and after, with a scanner finding added, removed and changed rows" width="100%">
3
+ </p>
4
+
5
+ <p align="center">
6
+ <a href="https://github.com/ChetanChirag24/rift/actions/workflows/ci.yml"><img alt="CI" src="https://github.com/ChetanChirag24/rift/actions/workflows/ci.yml/badge.svg"></a>
7
+ <img alt="Python 3.10+" src="https://img.shields.io/badge/python-3.10%2B-3776AB?logo=python&logoColor=white">
8
+ <img alt="Powered by DuckDB" src="https://img.shields.io/badge/powered%20by-DuckDB-FFF000?logo=duckdb&logoColor=black">
9
+ <img alt="Formats" src="https://img.shields.io/badge/formats-CSV%20%C2%B7%20Parquet%20%C2%B7%20JSON-22d3ee">
10
+ <img alt="Git aware" src="https://img.shields.io/badge/git-aware-F05032?logo=git&logoColor=white">
11
+ <img alt="MIT License" src="https://img.shields.io/badge/license-MIT-a78bfa">
12
+ </p>
13
+
14
+ <p align="center"><b>Think <code>git diff</code>, but for tables.</b><br>
15
+ Point rift at two versions of a dataset and it tells you which rows were <b>added</b>, which were <b>removed</b>, and exactly which values <b>changed</b>, all on your own machine.</p>
16
+
17
+ ---
18
+
19
+ ## Why rift?
20
+
21
+ Code changes get reviewed line by line. **Data changes usually don't.** A CSV export is regenerated, a SQL model is refactored, a price list is updated, and the only check is "the row count looks about right."
22
+
23
+ rift shows you what actually changed:
24
+
25
+ - πŸ” **Row-level answers.** Not just "3 rows differ": you see *which* rows and *which* columns, with before β†’ after values.
26
+ - 🏠 **Local-first.** Your data never leaves your laptop. No server, no account, no upload.
27
+ - ⚑ **Fast.** Both files are loaded into an in-memory [DuckDB](https://duckdb.org) database and compared with SQL, so millions of rows are fine.
28
+ - 🌿 **Git-aware.** Compare a file with any earlier commit: `rift HEAD~1:data.csv data.csv -k id`.
29
+ - πŸ€– **CI-ready.** Exits like `diff`: `0` = identical, `1` = different, `2` = error. Add `--json` for machine-readable output.
30
+
31
+ ## See it in action
32
+
33
+ <p align="center">
34
+ <img src="https://raw.githubusercontent.com/ChetanChirag24/rift/719315b1c043077b447e6c6ff89f3b1db84befa5/docs/assets/terminal.svg" alt="Terminal: rift compares data/customers.csv with its previous git commit and prints added, removed and changed rows" width="100%">
35
+ </p>
36
+
37
+ ## How it works
38
+
39
+ <p align="center">
40
+ <img src="https://raw.githubusercontent.com/ChetanChirag24/rift/719315b1c043077b447e6c6ff89f3b1db84befa5/docs/assets/how-it-works.svg" alt="Diagram: two files (one read from git) flow into a DuckDB cube which joins them on the key, then out to a rich table, JSON and an exit code" width="100%">
41
+ </p>
42
+
43
+ 1. **Load.** rift reads each side into DuckDB. Either side can be a CSV, Parquet or JSON file, or a version of a file from git history (`REV:path`).
44
+ 2. **Check the key.** The key column(s) you pass must exist on both sides and be unique and non-empty, so every row has exactly one partner.
45
+ 3. **Compare.** rift joins the two tables on the key:
46
+ - keys only in *after* β†’ **added**
47
+ - keys only in *before* β†’ **removed**
48
+ - keys on both sides where any column `IS DISTINCT FROM` its partner β†’ **changed** (so `NULL` vs `NULL` correctly counts as equal)
49
+ 4. **Report.** You get a colour table in the terminal, JSON with `--json`, and an exit code your scripts can act on. Added, removed and retyped columns are reported too.
50
+
51
+ ---
52
+
53
+ ## Real-world examples
54
+
55
+ ### 🧾 "Did last night's CRM export silently drop customers?"
56
+
57
+ Your finance dashboard is fed by a nightly export. Today's revenue number looks odd.
58
+
59
+ ```bash
60
+ rift exports/2026-11-15/customers.csv exports/2026-11-16/customers.csv -k customer_id
61
+ ```
62
+
63
+ ```text
64
+ Removed -1 ← Initech vanished from the export
65
+ Changed ~2 ← Globex: status active β†’ churned, revenue 8500 β†’ 0
66
+ ```
67
+
68
+ Two minutes later you know whether it's a real churn or a broken export.
69
+
70
+ ### πŸ› οΈ "I refactored a SQL model. Is the output exactly the same?"
71
+
72
+ You rewrote a 400-line query to make it faster. The output should not change at all.
73
+
74
+ ```bash
75
+ rift old_model.parquet new_model.parquet -k order_id --ignore _loaded_at
76
+ echo $? # 0 β†’ identical, ship it. 1 β†’ look at the diff first.
77
+ ```
78
+
79
+ `--ignore` skips columns that are *supposed* to differ, like load timestamps.
80
+
81
+ ### πŸ›’ "Which prices am I about to change in the store?"
82
+
83
+ Merchandising sent a new price sheet. Before you upload it, check exactly what it changes:
84
+
85
+ ```bash
86
+ rift prices_live.csv prices_new.csv -k sku --summary
87
+ ```
88
+
89
+ ```text
90
+ Changed values by column
91
+ column rows changed
92
+ price 412
93
+ sale_price 37
94
+ ```
95
+
96
+ 412 price changes when you expected 40? Better to catch that now than after customers do.
97
+
98
+ ### πŸ€– "Why did my model's accuracy drop between training runs?"
99
+
100
+ Diff the feature tables from both runs, keyed on a composite key, and pipe the JSON into `jq`:
101
+
102
+ ```bash
103
+ rift features_run41.parquet features_run42.parquet -k user_id,event_date --json \
104
+ | jq '.column_change_counts'
105
+ ```
106
+
107
+ ```json
108
+ { "avg_session_minutes": 120394, "country": 3 }
109
+ ```
110
+
111
+ One feature changed for 120k users. That's your bug.
112
+
113
+ ### 🌍 Multi-region tables with composite keys
114
+
115
+ The same `id` can exist in several regions? Combine columns into one key:
116
+
117
+ ```bash
118
+ rift before.csv after.csv -k region,id
119
+ ```
120
+
121
+ ---
122
+
123
+ ## Works with Git
124
+
125
+ Data files live in git repos too, and rift treats git history as a first-class source.
126
+
127
+ ### Compare with any commit, branch or tag
128
+
129
+ Use git's own `REV:path` syntax (paths are relative to the repo root, or start with `./` for the current directory):
130
+
131
+ ```bash
132
+ rift HEAD~1:data/customers.csv data/customers.csv -k customer_id # what did my last commit change?
133
+ rift main:data/customers.csv data/customers.csv -k customer_id # what does my branch change vs main?
134
+ rift v1.0:exports/users.parquet v2.0:exports/users.parquet -k id # between two releases
135
+ ```
136
+
137
+ ### Use rift as your `git difftool`
138
+
139
+ ```bash
140
+ git config --global difftool.rift.cmd 'rift "$LOCAL" "$REMOTE" --key id'
141
+ git difftool -y -t rift HEAD~1 -- data/customers.csv
142
+ ```
143
+
144
+ ### Catch data changes in pull requests
145
+
146
+ ```mermaid
147
+ flowchart LR
148
+ A[πŸ‘©β€πŸ’» Pull request<br/>changes data/] --> B[βš™οΈ CI runs<br/>rift main:data.csv data.csv]
149
+ B -->|exit 0| C[βœ… Data unchanged<br/>merge freely]
150
+ B -->|exit 1| D[⚠️ Data changed<br/>diff printed in the log]
151
+ B -->|exit 2| E[❌ Bad key or file<br/>fix the input]
152
+ D --> F[πŸ‘€ Reviewer checks<br/>the row-level diff]
153
+ F --> C
154
+ ```
155
+
156
+ A minimal GitHub Actions job:
157
+
158
+ ```yaml
159
+ name: data-diff
160
+ on: pull_request
161
+
162
+ jobs:
163
+ rift:
164
+ runs-on: ubuntu-latest
165
+ steps:
166
+ - uses: actions/checkout@v7
167
+ with:
168
+ fetch-depth: 0 # rift needs the main branch history
169
+ - uses: actions/setup-python@v7
170
+ with:
171
+ python-version: "3.12"
172
+ - run: pip install riftdiff
173
+ - name: Show what this PR changes in the data
174
+ run: rift origin/main:data/customers.csv data/customers.csv -k customer_id
175
+ continue-on-error: true # report the diff without failing the build
176
+ ```
177
+
178
+ Remove `continue-on-error` if any data change should block the merge until someone signs off.
179
+
180
+ ---
181
+
182
+ ## Install
183
+
184
+ ```bash
185
+ pip install riftdiff
186
+ ```
187
+
188
+ Or from source:
189
+
190
+ ```bash
191
+ git clone https://github.com/ChetanChirag24/rift.git && cd rift
192
+ python3 -m venv .venv && source .venv/bin/activate
193
+ pip install -e '.[dev]'
194
+ pytest
195
+ ```
196
+
197
+ > **macOS + iCloud Desktop/Documents:** iCloud can mark files inside `.venv` as hidden, and Python then skips the `.pth` file that makes the editable install work (`ModuleNotFoundError: No module named 'riftdiff'`). Create the venv as `.venv.nosync` (iCloud ignores `*.nosync`) and symlink it: `python3 -m venv .venv.nosync && ln -s .venv.nosync .venv`.
198
+
199
+ ## Usage
200
+
201
+ ```
202
+ rift BEFORE AFTER --key COLUMN [options]
203
+ ```
204
+
205
+ `BEFORE` and `AFTER` are file paths or git revisions (`HEAD~1:path/to/file.csv`).
206
+
207
+ | Option | What it does |
208
+ | --- | --- |
209
+ | `-k, --key COLUMN` | Column that identifies a row. Repeat it or comma-separate it for a composite key: `-k region,id`. |
210
+ | `-i, --ignore COLUMN` | Leave a column out of the comparison, e.g. `-i updated_at`. Repeatable / comma-separated. |
211
+ | `-n, --limit N` | Show at most N rows per section (default 10; `0` shows all). Counts are always exact. |
212
+ | `--summary` | Print only the counts, schema changes and per-column totals. |
213
+ | `--json` | Print the full diff as JSON for scripts and CI. |
214
+ | `--version` | Print the version. |
215
+
216
+ **Supported formats:** CSV/TSV (`.csv`, `.tsv`, `.txt`), Parquet (`.parquet`, `.pq`) and JSON (`.json`, `.jsonl`, `.ndjson`), optionally compressed with `.gz` or `.zst`. The two sides may be different formats.
217
+
218
+ ### What rift reports
219
+
220
+ | | Meaning |
221
+ | --- | --- |
222
+ | 🟒 **Added** | Keys that exist only in `AFTER` |
223
+ | πŸ”΄ **Removed** | Keys that exist only in `BEFORE` |
224
+ | 🟑 **Changed** | Keys on both sides where at least one value differs, with before β†’ after for each changed column |
225
+ | πŸ“Š **By column** | How many rows changed in each column |
226
+ | 🧬 **Schema** | Columns added, removed, or whose type changed. Retyped columns are compared by their text form, so `10` and `"10"` count as equal. |
227
+
228
+ rift refuses keys that are missing, empty or not unique, because a diff on those would be misleading.
229
+
230
+ ### Exit status
231
+
232
+ | Code | Meaning |
233
+ | --- | --- |
234
+ | `0` | No differences |
235
+ | `1` | Differences found |
236
+ | `2` | Error (missing file, bad key, unreadable data…) |
237
+
238
+ ## Python API
239
+
240
+ ```python
241
+ from riftdiff import diff
242
+
243
+ result = diff("HEAD~1:data/orders.csv", "data/orders.parquet",
244
+ key_column=["region", "order_id"],
245
+ ignore_columns=["updated_at"], limit=100)
246
+
247
+ result.summary() # {'rows_before': ..., 'rows_added': ..., ...}
248
+ result.added # list of row dicts
249
+ result.removed # list of row dicts
250
+ for change in result.changed:
251
+ print(change.key, change.changes) # {'region': 'EU', 'order_id': 7} {'total': (120, 135)}
252
+ result.has_differences # bool
253
+ result.to_dict() # JSON-ready dict
254
+ ```
255
+
256
+ Problems with the input raise `riftdiff.RiftError` with a readable message.
257
+
258
+ ## Try it now
259
+
260
+ The repo includes a small example:
261
+
262
+ ```bash
263
+ rift examples/customers_before.csv examples/customers_after.csv -k customer_id
264
+ ```
265
+
266
+ ## License
267
+
268
+ [MIT](https://github.com/ChetanChirag24/rift/blob/719315b1c043077b447e6c6ff89f3b1db84befa5/LICENSE)
@@ -0,0 +1,55 @@
1
+ [build-system]
2
+ requires = ["setuptools>=77.0"]
3
+ build-backend = "setuptools.build_meta"
4
+
5
+ [project]
6
+ name = "riftdiff"
7
+ dynamic = ["version"]
8
+ description = "Local-first data diff, powered by DuckDB. See exactly what changed before you merge."
9
+ readme = "README.md"
10
+ requires-python = ">=3.10"
11
+ license = "MIT"
12
+ license-files = ["LICENSE"]
13
+ authors = [{ name = "Chetan Chirag" }]
14
+ keywords = ["diff", "data", "csv", "parquet", "duckdb", "data-quality", "git"]
15
+ classifiers = [
16
+ "Development Status :: 4 - Beta",
17
+ "Environment :: Console",
18
+ "Intended Audience :: Developers",
19
+ "Operating System :: OS Independent",
20
+ "Programming Language :: Python :: 3",
21
+ "Programming Language :: Python :: 3 :: Only",
22
+ "Programming Language :: Python :: 3.10",
23
+ "Programming Language :: Python :: 3.11",
24
+ "Programming Language :: Python :: 3.12",
25
+ "Programming Language :: Python :: 3.13",
26
+ "Programming Language :: Python :: 3.14",
27
+ "Topic :: Database",
28
+ "Topic :: Software Development :: Quality Assurance",
29
+ ]
30
+ dependencies = [
31
+ "duckdb>=0.10.0",
32
+ "rich>=13.0.0",
33
+ ]
34
+
35
+ [project.urls]
36
+ Homepage = "https://github.com/ChetanChirag24/rift"
37
+ Source = "https://github.com/ChetanChirag24/rift"
38
+ Issues = "https://github.com/ChetanChirag24/rift/issues"
39
+ Changelog = "https://github.com/ChetanChirag24/rift/releases"
40
+
41
+ [project.optional-dependencies]
42
+ dev = ["pytest>=7"]
43
+
44
+ [project.scripts]
45
+ rift = "riftdiff.cli:main"
46
+
47
+ [tool.setuptools.dynamic]
48
+ version = { attr = "riftdiff.__version__" }
49
+
50
+ [tool.setuptools.packages.find]
51
+ where = ["src"]
52
+
53
+ [tool.pytest.ini_options]
54
+ testpaths = ["tests"]
55
+ pythonpath = ["src"]
@@ -0,0 +1,4 @@
1
+ [egg_info]
2
+ tag_build =
3
+ tag_date = 0
4
+
@@ -0,0 +1,7 @@
1
+ """rift: local-first data diff, powered by DuckDB."""
2
+
3
+ __version__ = "0.1.0"
4
+
5
+ from .engine import DiffResult, RiftError, RowChange, diff # noqa: E402
6
+
7
+ __all__ = ["diff", "DiffResult", "RowChange", "RiftError", "__version__"]
@@ -0,0 +1,5 @@
1
+ import sys
2
+
3
+ from .cli import main
4
+
5
+ sys.exit(main())