string_selection_resolver

Utilities for resolving string-based selections to event indices.

class graphnet.data.utilities.string_selection_resolver.StringSelectionResolver(dataset, index_column, seed, use_cache)

Bases: Logger

Resolve string-based selection to event indices.

String-based selection, using in DatasetConfig, is a very flexible way of defining event selections. Below we show an example of a very involved event selection, which should cover most standard event selections currently in use with graphnet.

```yml # dataset/config.yml selection:

test:

• 50000 random events ~ event_no % 5 == 0 & abs(pid) == 12

• 50000 random events ~ event_no % 5 == 0 & abs(pid) == 14

• 50000 random events ~ event_no % 5 == 0 & abs(pid) == 16

• 50000 random events ~ event_no % 5 == 0 & abs(pid) == 13

• 50000 random events ~ event_no % 5 == 0 & abs(pid) == 1

validation:

• 10000 random events ~ event_no % 5 == 1 & abs(pid) == 12

• 10000 random events ~ event_no % 5 == 1 & abs(pid) == 14

• 10000 random events ~ event_no % 5 == 1 & abs(pid) == 16

• 10000 random events ~ event_no % 5 == 1 & abs(pid) == 13

• 10000 random events ~ event_no % 5 == 1 & abs(pid) == 1

train:

• 10000 random events ~ event_no % 5 > 1 & abs(pid) == 12

• 10000 random events ~ event_no % 5 > 1 & abs(pid) == 14

• 10000 random events ~ event_no % 5 > 1 & abs(pid) == 16

• 10000 random events ~ event_no % 5 > 1 & abs(pid) == 13

• 10000 random events ~ event_no % 5 > 1 & abs(pid) == 1

```

Construct StringSelectionResolver.

Parameters:

• dataset (Dataset)

• index_column (str)

• seed (int | None)

• use_cache (bool)

resolve(selection)

Resolve selection as string to list of indicies.

Selections are expected to have pandas.DataFrame.query-compatible syntax, e.g., ` "event_no % 5 > 0" ` Selections may also specify a fixed number of events to randomly sample, e.g., ` "10000 random events ~ event_no % 5 > 0" "20% random events ~ event_no % 5 > 0" `

Return type:

List[int]

Parameters:

selection (str)