# Summary

## Problem statement

The `Resource` class is also a [Context Manager](https://docs.python.org/3/reference/datamodel.html#context-managers).  That is, it implements the
`__enter()__` and `__exit()__` methods to allow the use of
`with Resource(...)` statements.

Prior to this PR, there was no limit on nesting `with` statements on
the same `Resource`, but this caused problems because while the second
`__enter()__` allowed the `Resource` to already be open, the first
`__exit()__` would `close()` the Resource while the higher level context
would expect it to still be open.

This would cause errors like "ValueError: I/O operation on closed file",
or the iterator would appear to start from part way through a file rather
than at the start of the file, and other similar behaviour depending on
the exact locations of the nested functions.

This was made more complex because these `with` statements were often
far removed from each other in the code, hidden behind iterators driven
by generators, etc. They also could have different behaviour depending on
number of rows read, the type of Resource (local file vs inline, etc.),
the different steps in a pipeline, etc. etc.

All this meant that the problem was rare, hard to reduce down to an
obvious reproduction case, and not realistic to expect developers to
understand while developing new functionality.

## Solution

This PR prevents nested contexts being created by throwing an exception
when the second, nested, `with` is attempted.  This means that code that
risks these issues can be quickly identified and resolved during development.

The best way to resolve it is to use `Resource.to_copy()` to copy so that
the nested `with` is acting on an independent view of the same Resource,
which is likely what is intended in most cases anyway.

This PR also updates a number of the internal uses of `with` to work
on a copy of the Resource they are passed so that they are independent
of any external code and what it might have done with the Resource prior
to the library methods being called.

## Breaking Change

This is technically a breaking change as any external code that was
developed using nested `with` statements - possibly deliberately, but
more likely unknowingly not falling into the error cases - will have to
be updated to use `to_copy()` or similar.

However, the library functions have all been updated in a way that
doesn't change their signature or their expected behaviour as documented
by the unit tests. All pre-existing unit tests pass with no changes,
and added unit tests for the specific updated behaviour do not require
any unusual constructs.  It is still possible that some undocumented
and untested side effect behaviours are different than before and any
code relying on those may also be affected (e.g. `to_petl()` iterators are now independent rather than causing changes in each other)

So it is likely that very few actual impacts will occur in real world
code, and the exception thrown does it's best to explain the issue
and suggest resolutions.

# Tests

- All existing unit tests run and pass unchanged
- New unit tests were added to cover the updated behaviour
    - These unit tests were confirmed to fail without the updates
      in this PR (where appropriate).
    - These unit tests now pass with the updated code.
- The original script that identified the issue in frictionlessdata#1622 was run and
  now gives the correct result (all rows appropriately converted and
  saved to file)

# Summary

This fixes formatting issues identified by the linter.  No functional changes are included.

# Tests

- hatch run lint passes
- hatch run test passes (and coverage is sufficient)

# Summary

The CI tests identified some issues that don't show up on a normal test run.  This commit fixes those issues.

It also highlighted that there were numerous areas
that didn't have sufficient test coverage for the case
that the caller had already opened the resource.

The indexer has some notable changes, but the biggest
area affected is the parsers when writing
from an already opened source.  This commit adds
unit tests for the index and all the parser formats for this case,
and fixes the code to support the lack of nested contexts.

# Tests

- Setup the required databases for CI by copying the commands in the github actions
- Run `hatch run +py=3.11 ci:test` and ensure all tests pass and coverage remains sufficient
- Run `hatch run test` in case it is different and ensure all tests pass and coverage remains sufficient

This also means that all linting etc. has been run too.