exp show: Add --drop and --keep args

[daavoo]

• ❗ I have followed the Contributing to DVC checklist.

• 📖 If this PR requires documentation updates, I have created a separate PR (or issue, at least) in dvc.org and linked it here.

---

• Removed --include-metrics / --include-params and
  --exclude-metrics / --exclude-params.

• Removed --no-timestamp (-> --drop Created)

• --drop and --keep operate directly on the table columns.
  Support regex patterns.

• --keep does not perform any filtering.
  It's only used to specify columns to keep despite the other filtering.
  Previous --include behavior can be replicated with negative lookahead regex.

Closes #7079
Closes #7080
Closes #7083
Pre-requisite for #6434

---

Example:

`dvc exp show`

| Experiment              | Created      | loss    | accuracy   | train.batch_size   | train.hidden_units   | train.dropout   | train.num_epochs   | train.lr   | train.conv_activation   |
|-------------------------|--------------|---------|------------|--------------------|----------------------|-----------------|--------------------|------------|-------------------------|
| workspace               | -            | 0.31941 | 0.8997     | 128                | 64                   | 0.3             | 10                 | 0.01       | relu                    |
| main                    | Sep 14, 2021 | 0.26484 | 0.9038     | 128                | 64                   | 0.4             | 10                 | 0.001      | relu                    |
| ├── d599d35 [exp-4f1f0] | 06:12 PM     | 0.31941 | 0.8997     | 128                | 64                   | 0.3             | 10                 | 0.01       | relu                    |
| ├── c6d3ac6 [exp-eec4c] | 06:10 PM     | 0.24435 | 0.9135     | 128                | 64                   | 0.3             | 10                 | 0.001      | relu                    |
| └── 0d5e1ae [exp-25fae] | 06:03 PM     | 0.26133 | 0.9062     | 128                | 64                   | 0.4             | 10                 | 0.001      | relu                    |

`dvc exp show --only-changed`

| Experiment              | Created      | loss    | accuracy   | train.dropout   | train.lr   |
|-------------------------|--------------|---------|------------|-----------------|------------|
| workspace               | -            | 0.31941 | 0.8997     | 0.3             | 0.01       |
| main                    | Sep 14, 2021 | 0.26484 | 0.9038     | 0.4             | 0.001      |
| ├── d599d35 [exp-4f1f0] | 06:12 PM     | 0.31941 | 0.8997     | 0.3             | 0.01       |
| ├── c6d3ac6 [exp-eec4c] | 06:10 PM     | 0.24435 | 0.9135     | 0.3             | 0.001      |
| └── 0d5e1ae [exp-25fae] | 06:03 PM     | 0.26133 | 0.9062     | 0.4             | 0.001      |

`dvc exp show --only-changed --keep train.batch_size`

| Experiment              | Created      | loss    | accuracy   | train.batch_size   | train.dropout   | train.lr   |
|-------------------------|--------------|---------|------------|--------------------|-----------------|------------|
| workspace               | -            | 0.31941 | 0.8997     | 128                | 0.3             | 0.01       |
| main                    | Sep 14, 2021 | 0.26484 | 0.9038     | 128                | 0.4             | 0.001      |
| ├── d599d35 [exp-4f1f0] | 06:12 PM     | 0.31941 | 0.8997     | 128                | 0.3             | 0.01       |
| ├── c6d3ac6 [exp-eec4c] | 06:10 PM     | 0.24435 | 0.9135     | 128                | 0.3             | 0.001      |
| └── 0d5e1ae [exp-25fae] | 06:03 PM     | 0.26133 | 0.9062     | 128                | 0.4             | 0.001      |

`dvc exp show --drop train.*`

| Experiment              | Created      | loss    | accuracy   |
|-------------------------|--------------|---------|------------|
| workspace               | -            | 0.31941 | 0.8997     |
| main                    | Sep 14, 2021 | 0.26484 | 0.9038     |
| ├── d599d35 [exp-4f1f0] | 06:12 PM     | 0.31941 | 0.8997     |
| ├── c6d3ac6 [exp-eec4c] | 06:10 PM     | 0.24435 | 0.9135     |
| └── 0d5e1ae [exp-25fae] | 06:03 PM     | 0.26133 | 0.9062     |

`dvc exp show --drop train.* --keep train.dropout`

| Experiment              | Created      | loss    | accuracy   | train.dropout   |
|-------------------------|--------------|---------|------------|-----------------|
| workspace               | -            | 0.31941 | 0.8997     | 0.3             |
| main                    | Sep 14, 2021 | 0.26484 | 0.9038     | 0.4             |
| ├── d599d35 [exp-4f1f0] | 06:12 PM     | 0.31941 | 0.8997     | 0.3             |
| ├── c6d3ac6 [exp-eec4c] | 06:10 PM     | 0.24435 | 0.9135     | 0.3             |
| └── 0d5e1ae [exp-25fae] | 06:03 PM     | 0.26133 | 0.9062     | 0.4             |

`dvc exp show --only-changed --drop Created`

| Experiment              | loss    | accuracy   | train.dropout   | train.lr   |
|-------------------------|---------|------------|-----------------|------------|
| workspace               | 0.31941 | 0.8997     | 0.3             | 0.01       |
| main                    | 0.26484 | 0.9038     | 0.4             | 0.001      |
| ├── d599d35 [exp-4f1f0] | 0.31941 | 0.8997     | 0.3             | 0.01       |
| ├── c6d3ac6 [exp-eec4c] | 0.24435 | 0.9135     | 0.3             | 0.001      |
| └── 0d5e1ae [exp-25fae] | 0.26133 | 0.9062     | 0.4             | 0.001      |

For #7080

`dvc exp show --only-changed --drop Experiment --drop Created`

| loss    | accuracy   | train.dropout   | train.lr   |               
|---------|------------|-----------------|------------|
| 0.31941 | 0.8997     | 0.3             | 0.01       |
| 0.26484 | 0.9038     | 0.4             | 0.001      |
| 0.31941 | 0.8997     | 0.3             | 0.01       |
| 0.24435 | 0.9135     | 0.3             | 0.001      |
| 0.26133 | 0.9062     | 0.4             | 0.001      |

[pmrowla]

How does filtering with the new options work if there is a metric and parameter with the same name?

[karajan1001]

Cool, I don't like the old _filter_name and _filter_names.

[daavoo]

How does filtering with the new options work if there is a metric and parameter with the same name?

User need to indicate the pattern matching the final column. So if there is a shared name, user would need to --drop params.yaml:foo or --drop *foo

[dberenbaum]

Looks great! Nice simplification of the UI.

The big question is whether removing --include flags will make it too hard to filter the table. Now that we have --only-changed, the hope is that this will cover a lot that filtering automatically (along with --drop).

We haven't had much feedback recently on the table filtering, so let's try it and see what people think.

[daavoo]

Looks great! Nice simplification of the UI.

The big question is whether removing --include flags will make it too hard to filter the table. Now that we have --only-changed, the hope is that this will cover a lot that filtering automatically (along with --drop).

We haven't had much feedback recently on the table filtering, so let's try it and see what people think.

I was thinking that if we support regex matching instead of fnmatch there would be no need of --keep and users could perform similar behavior to previous --include.

I'm Ok with adding regex and just examples in docs. @dberenbaum WDYT?

[dberenbaum]

I was thinking that if we support regex matching instead of fnmatch there would be no need of --keep and users could perform similar behavior to previous --include.

I'm Ok with adding regex and just examples in docs. @dberenbaum WDYT?

I knew we had this discussion before 😄 : #5720 (comment).

No strong preference from me. Not sure if regex is less consistent with the rest of the dvc cli, but obviously it's more powerful, so 🤷 .

[pmrowla]

@daavoo this needs a rebase, exp commands have been split into their own files (so dvc/commands/experiments/show.py) since there was getting to be way too much stuff in commands/experiments.py

[daavoo]

@daavoo this needs a rebase, exp commands have been split into their own files (so dvc/commands/experiments/show.py) since there was getting to be way too much stuff in commands/experiments.py

Done. I had already done that one but this required a new one because of the parallel coordinates merge

[jorgeorpinel]

Nice. On first sight it's not clear to me what the difference between include and keep are but we'll worry about that in the docs PR. Just one Q: Does the order in which filters are given matter? I.e. if I keep something and then drop the same something what happens? I'd even try --keep foo --drop foo --keep foo P: and see what happens. Thanks

UPDATE: I guess it's already explained in iterative/dvc.org#3112 so nvmd.

[lgtm-com]

This pull request introduces 1 alert when merging 556a6db97d321d86cf56bfe8df8899e199aeb2f4 into 0a4958b - view on LGTM.com

new alerts:

• 1 for Unused import

[karajan1001]

LGTM, in the most parts.

[karajan1001]

How about remove to_remove and directly:

        for col in col_names:
            if not self.is_protected(col):
                self._keys.remove(col)
                self._columns.pop(col)

[karajan1001]

It merged automatically.

[daavoo]

#7299