Cobra

Please see https://github.com/opencobra/cobrapy/tree/0.9.1/release-notes/0.9.1.md for the full release notes.

Fixes

- `Model.compartment` is now a dynamic property fetching the
compartments from all metabolites therefore always
up-to-date. Assigning a dictionary to the same property updates the
internal dictionary of compartment descriptions. This change removes
the need for the check for missing compartments from
`validation.check_metabolite_compartment_formula`.
- Excessively long output of html representations in jupyter notebooks are now abbreviated [577](https://github.com/opencobra/cobrapy/pull/577).
- Reaction forward and reverse variables are no longer cached with those object. No visible effect but simplifies the code.
- Fix bug in summary methods when used with exchange reaction sas objective. [595](https://github.com/opencobra/cobrapy/pull/595).

New features

- `Model.objective_direction` is a new revertible property to set maximization / minimization using the context manager.
- Change output of `production_envelope` to wide data frame format [587](https://github.com/opencobra/cobrapy/pull/587). Also allow multiple carbon source reactions and better handling of zero-division exceptions.
- Enable summing lists of reactions, see [596](https://github.com/opencobra/cobrapy/pull/596)

Deprecated features

- `Model.get_metabolite_compartments` is deprecated (use
`Model.compartments` instead).
- `Reaction.get_compartments` is deprecated (use
`Reaction.compartments` instead).

Backwards incompatible changes

- The format of the dataframe `production_envelope` changed listing max and min on different columns instead of the same.

Fixes

- Guarantee that sampler._reproject always returns a feasible point
and will not attempt to reproject already feasible
points. [564](https://github.com/opencobra/cobrapy/pull/564)
- `Model.summary` no longer fails when calling after the model has
changed. Fixed by letting the summary function re-compute a solution
(default) or letting user supply a prior computed solution
object. [566](https://github.com/opencobra/cobrapy/pull/566)
- Metabolites must now have valid identifiers before being added to a
model or `ValueError` is raised.
- Fix use of underscores in key/value pairs in legacy sbml
notes. [547](https://github.com/opencobra/cobrapy/issues/547)

Backwards incompatible changes

- the Solution class no longer contains links progenitor model's
reactions and metabolites. Removed since it those can change after
the solution has been computed making them erroneous. This of course
implies that `Solution` constructor changes to:

def __init__(self, objective_value, status, fluxes, reduced_costs=None,
shadow_prices=None, **kwargs):

Fixes

- Fix bug related to inhomogeneous sampling
[558](https://github.com/opencobra/cobrapy/issues/558). During
flux sampling, store the nullspace rather than the entire projection
(N * N.T) which reduces the memory footprint since it uses memory in
the order of `2 * n_reaction * n_metabolites` instead of `4 *
n_reactions^2` (no change in
speed). [PR 556](https://github.com/opencobra/cobrapy/pull/556)

Fixes

- Tests no longer generates warnings about using deprecated functions.
- Gapfilling integrality thresholds now supported for all optlang solvers.

New features

- `Model.slim_optimize()` can be used perform optimization without
creating a solution. Can lead to significant speedup compared to
`Model.optimize` when repeatedly doing optimizations and only making
use of the objective value as avoiding the need to fetch all values
from the solver object.
- solution, model, metabolite and reaction now have html
representation so they give more informative prints in jupyter
notebooks.
- New convenience functions `cobra.flux_analysis.find_essential_genes` and
`cobra.flux_analysis.find_essential_reactions`.
- `Model.optimize` has new parameter `raise_error` to enable option to
get trigger exception if no feasible solution could be found.
- `str(reaction)` now gives the more useful reaction id and the
reaction string.

Deprecated features

- `str(reaction)` no longer gives `reaction.id`.

Fixes

- `cobra.flux_analysis.reaction.assess`
[was broken](https://github.com/opencobra/cobrapy/issues/537)
following the release of 0.6.0 and has now been fixed (and now with
unit tests).
- `production_envelope` failed when model C-source was formulated as
-> x instead of x <-. Fixed added option to guess the C-source by
taking the medium reaction with the highest input C flux.
- `model_to_pymatbridge` needs scipy and that's correctly handled now.

New features
- `flux_variability_analysis` now has the `pfba_factor` parameter
which enables the inclusion of a constraint on the max sum of
absolute fluxes when doing FVA.

Deprecated features

- `cobra.flux_analysis.reaction.assess_{precursors,products}` were
essentially copies of each other and have been merged to
`cobra.flux_analysis.reaction.assess_component`