Module tf.core.api
The core API of TF.
It provides methods to navigate nodes and edges and lookup features.
Expand source code Browse git
"""
# The core API of TF.
It provides methods to navigate nodes and edges and lookup features.
"""
import collections
from textwrap import wrap, dedent
from .helpers import flattenToSet, console, fitemize, deepSize
from ..advanced.helpers import dm
from .files import unexpanduser as ux
from .nodes import Nodes
from .locality import Locality
from .nodefeature import NodeFeatures
from .edgefeature import EdgeFeatures
from .computed import Computeds
from .text import Text
from ..parameters import OTYPE, OSLOTS
from ..search.search import Search
from ..core.timestamp import SILENT_D, DEEP, silentConvert
API_REFS = dict(
AllComputeds=("Computed", "computedall", "computed-data"),
AllEdges=("Features", "edgeall", "edge-features"),
AllFeatures=("Features", "nodeall", "node-features"),
C=("Computed", "computed", "computed-data"),
Call=("Computed", "computedall", "computed-data"),
Computed=("Computed", "computed", "computed-data"),
ComputedString=("Computed", "computedstr", "computed-data"),
Cs=("Computed", "computedstr", "computed-data"),
E=("Features", "edge", "edge-features"),
Eall=("Features", "edgeall", "edge-features"),
Edge=("Features", "edge", "edge-features"),
EdgeString=("Features", "edgestr", "edge-features"),
Es=("Features", "edgestr", "edge-features"),
F=("Features", "node", "node-features"),
Fall=("Features", "nodeall", "node-features"),
Feature=("Features", "node", "node-features"),
FeatureString=("Features", "nodestr", "node-features"),
Fs=("Features", "nodestr", "node-features"),
L=("Locality", "locality", "locality"),
Locality=("Locality", "locality", "locality"),
N=("Nodes", "nodes", "navigating-nodes"),
Nodes=("Nodes", "nodes", "navigating-nodes"),
S=("Search", "search", "search"),
Search=("Search", "search", "search"),
T=("Text", "text", "text"),
TF=("Fabric", "fabric", "loading"),
Text=("Text", "text", "text"),
)
class Api:
def __init__(self, TF):
self.TF = TF
self.ignored = tuple(sorted(TF.featuresIgnored))
"""Which features were found but ignored.
Features are ignored if the feature is also present in another location
that is loaded later.
"""
TF.ignored = self.ignored
self.F = NodeFeatures()
self.Feature = self.F
self.E = EdgeFeatures()
self.Edge = self.E
self.C = Computeds()
self.Computed = self.C
tmObj = TF.tmObj
TF.silentOn = tmObj.silentOn
TF.silentOff = tmObj.silentOff
TF.isSilent = tmObj.isSilent
TF.setSilent = tmObj.setSilent
TF.info = tmObj.info
TF.warning = tmObj.warning
TF.error = tmObj.error
TF.cache = tmObj.cache
TF.reset = tmObj.reset
TF.indent = tmObj.indent
"""All messages produced during the feature loading process.
It also shows the messages that have been suppressed due to the `silent`
parameter.
"""
TF.ensureLoaded = self.ensureLoaded
TF.makeAvailableIn = self.makeAvailableIn
TF.footprint = self.footprint
setattr(self, "FeatureString", self.Fs)
setattr(self, "EdgeString", self.Es)
setattr(self, "ComputedString", self.Cs)
setattr(self, "AllFeatures", self.Fall)
setattr(self, "AllEdges", self.Eall)
setattr(self, "AllComputeds", self.Call)
setattr(self, "loadLog", self.isLoaded)
def Fs(self, fName, warn=True):
"""Get the node feature sub API.
If feature name is not a valid python identifier,
or if you do not know its name in advance,
you can not use `F.feature`, but you should use
`Fs(feature)`.
Parameters
----------
fName: string
The name of the feature.
warn: boolean, optional `True`
Whether to warn if the feature is not loaded.
Returns
-------
The feature API, or `None` if the feature is not loaded.
"""
if not hasattr(self.F, fName):
if warn:
self.TF.error(f'Node feature "{fName}" not loaded')
return None
return getattr(self.F, fName)
def Es(self, fName, warn=True):
"""Get the edge feature sub API.
If feature name is not a valid python identifier,
or if you do not know its name in advance,
you can not use `E.feature`, but you should use
`Es(feature)`.
Parameters
----------
fName: string
The name of the feature.
warn: boolean, optional `True`
Whether to warn if the feature is not loaded.
Returns
-------
The feature API, or `None` if the feature is not loaded.
"""
if not hasattr(self.E, fName):
if warn:
self.TF.error(f'Edge feature "{fName}" not loaded')
return None
return getattr(self.E, fName)
def Cs(self, fName, warn=True):
"""Get the computed data sub API.
If component name is not a valid python identifier,
or if you do not know its name in advance,
you can not use `C.component`, but you should use
`Cs(component)`.
Parameters
----------
fName: string
The name of the feature.
warn: boolean, optional `True`
Whether to warn if the feature is not loaded.
Returns
-------
The feature API, or `None` if the feature is not loaded.
"""
if not hasattr(self.C, fName):
if warn:
self.TF.error(f'Computed feature "{fName}" not loaded')
return None
return getattr(self.C, fName)
def Fall(self, warp=True):
"""Returns a sorted list of all usable, loaded node feature names.
Parameters
----------
warp: boolean, optional True
Whether to include the warp features, i.e. `otype`.
"""
return sorted(x[0] for x in self.F.__dict__.items() if warp or x[0] != OTYPE)
def Eall(self, warp=True):
"""Returns a sorted list of all usable, loaded edge feature names.
Parameters
----------
warp: boolean, optional True
Whether to include the warp features, i.e. `oslots`.
"""
return sorted(x[0] for x in self.E.__dict__.items() if warp or x[0] != OSLOTS)
def Call(self):
"""Returns a sorted list of all usable, loaded computed data names."""
return sorted(x[0] for x in self.C.__dict__.items())
def isLoaded(
self, features=None, pretty=True, valueType=True, path=False, meta="description"
):
"""Show information about loaded features.
Parameters
----------
features: iterable | string, optional None
The features to get info for.
If absent or None: all features seen by TF.
If a string, it is a comma and/or space spearated list of feature names.
Otherwise the items of the iterable are feature names.
pretty: boolean, optional True
If True, it prints an overview of all features seen by TF with
information about kind, type, source location and loaded status.
The amount of information printed can be tweaked by other parameters.
Otherwise, it returns this information as a dict.
valueType: boolean, optional True
Only relevant if `pretty=True`: whether to print the value type of
the values in the feature file.
path: boolean, optional True
Only relevant if `pretty=True`: whether to print the path name of
the feature file.
meta: string|list|boolean, optional "description"
Only relevant if `pretty=True`: controls what metadata of the feature
should be printed.
If it is None, False, or the empty string or empty list:
no metadata will be printed.
It it is the boolean value True: all metadata will be printed.
If it is a list of key names or a string with key names
separated by white space and/or commas, only these metadata keys
will be printed.
Returns
-------
dict of dict
The features are keys, the value per feature is None or a dict with the
following information:
`None` if the feature is not loaded.
If the feature is loaded:
* `kind`: `node`, `edge`, `config`, `computed`;
* `type` is the type of values: `int`, or `str` or `""`;
* `edgeValues`: if an edge feature it indicates whether
the edges have values. Otherwise `None`.
* `meta`: dictionary containing the metadata of the feature
If `pretty`, nothing is returned, but the dict is pretty printed.
"""
fNames = list(self.TF.features) if features is None else fitemize(features)
info = {}
for fName in fNames:
fMeta = {}
fType = None
edgeValues = None
fSource = None
hasInfo = True
if fName in self.TF.features:
fObj = self.TF.features[fName]
fSource = ux(fObj.dirName)
fMeta = fObj.metaData
fType = fMeta.get("valueType", "")
fMeta = {k: v for (k, v) in fMeta.items() if k != "valueType"}
isLoadedF = hasattr(self.F, fName)
isLoadedE = hasattr(self.E, fName)
if isLoadedF or isLoadedE:
if isLoadedF:
fKind = "node"
elif isLoadedE:
fKind = "edge"
flObj = getattr(self.E, fName)
edgeValues = False if fName == "oslots" else flObj.doValues
elif (
fName.startswith("__")
and fName.endswith("__")
and hasattr(self.C, fName.strip("_"))
):
fKind = "computed"
elif fName in self.TF.features:
if fObj.isConfig:
fKind = "config"
else:
hasInfo = False
else:
hasInfo = False
info[fName] = (
dict(
kind=fKind,
type=fType,
meta=fMeta,
source=fSource,
edgeValues=edgeValues,
)
if hasInfo
else None
)
if pretty:
for (fName, fInfo) in sorted(info.items()):
if fInfo is None:
kind = "NOT LOADED"
kind = f" {kind:<10}"
fSource = ""
metaRep = ""
heading = f"{fName:<20}{kind}{fSource}"
else:
fKind = fInfo["kind"]
fMeta = fInfo.get("meta", {})
fType = fInfo.get("type", "")
fSource = fInfo.get("source", "") if path else ""
fSource = f" {fSource}" if fSource else ""
fEV = fInfo.get("edgeValues", "")
if valueType:
kind = (
f"node ({fType})"
if fKind == "node"
else f"edge ({fType})"
if fKind == "edge" and fEV
else "edge"
if fKind == "edge"
else f"{fKind}"
)
kind = f" {kind:<10}" if kind else ""
else:
kind = ""
if meta is True:
metaKeys = sorted(fMeta.keys())
metaInfo = fMeta
elif not meta:
metaInfo = {}
else:
metaKeys = fitemize(meta)
metaInfo = {k: fMeta[k] for k in metaKeys if k in fMeta}
heading = f"{fName:<20}{kind}{fSource}"
metaRep = ""
indent = " " * (len(heading) + 1)
if metaInfo:
if len(metaKeys) == 1:
value = metaInfo.get(metaKeys[0], "")
value = "\n".join(
wrap(value, width=80, subsequent_indent=indent)
)
metaRep = f" {value}" if value else ""
else:
indent = " " * 21
for k in metaKeys:
value = metaInfo.get(k, "")
value = "\n".join(
wrap(
value,
width=80,
subsequent_indent=f"\t{indent} ",
)
)
metaRep += f"\n\t{k:<20} = {value}"
msg = f"{heading}{metaRep}"
print(msg)
return None
return info
def makeAvailableIn(self, scope):
"""Exports members of the API to the global namespace.
Only the members whose names start with a capital are exported.
If you are working with a single data source in your program, it is a bit
tedious to write the initial `TF.api.` or `A.api` all the time.
By this method you can avoid that.
!!! explanation "Longer names"
There are also longer names which can be used as aliases
to the single capital letters.
This might or might not improve the readability of your program.
short name | long name
--- | ---
`N` | `Nodes`
`F` | `Feature`
`Fs` | `FeatureString`
`Fall` | `AllFeatures`
`E` | `Edge`
`Es` | `EdgeString`
`Eall` `AllEdges`
`C` | `Computed`
`Cs` `ComputedString`
`Call` | `AllComputeds`
`L` | `Locality`
`T` | `Text`
`S` | `Search`
Parameters
----------
scope: dict
A dictionary into which the members of the core API will be inserted.
The only sensible choice is: `globals()`.
Returns
-------
tuple
A grouped list of API members that has been hoisted to the global
scope.
Notes
-----
!!! explanation "Why pass `globals()`?"
Although we know it should always be `globals()`, we cannot
define a function that looks into the `globals()` of its caller.
So we have to pass it on.
"""
for member in dir(self):
if "_" not in member and member[0].isupper():
scope[member] = getattr(self, member)
if member not in API_REFS:
console(f'WARNING: API member "{member}" not documented')
grouped = {}
for (member, (head, sub, ref)) in API_REFS.items():
grouped.setdefault(ref, {}).setdefault((head, sub), []).append(member)
# grouped
# node-features=>(Features, node)=>[F, ...]
docs = []
for (ref, groups) in sorted(grouped.items()):
chunks = []
for ((head, sub), members) in sorted(groups.items()):
chunks.append(" ".join(sorted(members, key=lambda x: (len(x), x))))
docs.append((head, ref, tuple(chunks)))
return docs
# docs
# (Features, node-features, ('F ...', ...))
def ensureLoaded(self, features):
"""Checks if features are loaded and if not loads them.
All features in question will be made available to the core API.
Parameters
----------
features: string | iterable of strings
It is a string containing space separated feature names,
or an iterable of feature names.
The feature names are just the names of `.tf` files
without directory information and without extension.
Returns
-------
set
The names of the features in question as a set of strings.
"""
F = self.F
E = self.E
TF = self.TF
warning = TF.warning
needToLoad = set()
loadedFeatures = set()
for fName in sorted(flattenToSet(features)):
fObj = TF.features.get(fName, None)
if not fObj:
warning(f'Cannot load feature "{fName}": not in dataset')
continue
if fObj.dataLoaded and (hasattr(F, fName) or hasattr(E, fName)):
loadedFeatures.add(fName)
else:
needToLoad.add(fName)
if len(needToLoad):
TF.load(
needToLoad,
add=True,
silent=DEEP,
)
loadedFeatures |= needToLoad
return loadedFeatures
def footprint(self, recompute=False, bySize=True):
"""Computes the memory footprint in RAM of the loaded TF data.
This includes the precomputed data.
Parameters
----------
recompute: boolean, optional False
The function looks first for earlier computed size data.
If that is found, it will be used, and no size computation will take place.
Unless this parameter is True.
If no earlier computed size data is found, sizes will be computed anyway.
bySize: boolean, optional True
Whether to sort the features by the size they occupy in RAM.
If False, the features will be sorted alphabetically.
"""
if hasattr(self, "sizes") and not recompute:
sizes = self.sizes
else:
TF = self.TF
features = TF.features
nFeatures = len(features)
sizes = {}
for ft in sorted(features):
console(f"\rcomputing size of {ft:<30}", newline=False)
data = features[ft].data
if data is None:
continue
nData = len(data)
sData = deepSize(data)
sizes[ft] = (nData, sData)
console(f'\r{"":>40}', newline=False)
self.sizes = sizes
material = ""
nFeatures = len(sizes)
totals = collections.Counter()
for (ft, (nData, sData)) in sorted(
sizes.items(),
key=(lambda x: (-x[1][1], x[0])) if bySize else lambda x: x[0],
):
material += f"{ft} | {nData:,} | {sData:,}\n"
totals["nData"] += nData
totals["sData"] += sData
material += f'TOTAL | {totals["nData"]:,} | {totals["sData"]:,}'
header = dedent(
f"""
# {nFeatures} features
feature | members | size in bytes
--- | --- | ---
"""
)
dm(header + material)
def addOtype(api):
setattr(api.F.otype, "all", tuple(o[0] for o in api.C.levels.data))
setattr(
api.F.otype, "support", dict(((o[0], (o[2], o[3])) for o in api.C.levels.data))
)
def addLocality(api):
api.L = Locality(api)
api.Locality = api.L
def addNodes(api):
api.N = Nodes(api)
api.Nodes = api.N
def addText(api):
api.T = Text(api)
api.Text = api.T
def addSearch(api, silent=SILENT_D):
silent = silentConvert(silent)
api.S = Search(api, silent)
api.Search = api.SFunctions
def addLocality(api)-
Expand source code Browse git
def addLocality(api): api.L = Locality(api) api.Locality = api.L def addNodes(api)-
Expand source code Browse git
def addNodes(api): api.N = Nodes(api) api.Nodes = api.N def addOtype(api)-
Expand source code Browse git
def addOtype(api): setattr(api.F.otype, "all", tuple(o[0] for o in api.C.levels.data)) setattr( api.F.otype, "support", dict(((o[0], (o[2], o[3])) for o in api.C.levels.data)) ) def addSearch(api, silent='auto')-
Expand source code Browse git
def addSearch(api, silent=SILENT_D): silent = silentConvert(silent) api.S = Search(api, silent) api.Search = api.S def addText(api)-
Expand source code Browse git
def addText(api): api.T = Text(api) api.Text = api.T
Classes
class Api (TF)-
Expand source code Browse git
class Api: def __init__(self, TF): self.TF = TF self.ignored = tuple(sorted(TF.featuresIgnored)) """Which features were found but ignored. Features are ignored if the feature is also present in another location that is loaded later. """ TF.ignored = self.ignored self.F = NodeFeatures() self.Feature = self.F self.E = EdgeFeatures() self.Edge = self.E self.C = Computeds() self.Computed = self.C tmObj = TF.tmObj TF.silentOn = tmObj.silentOn TF.silentOff = tmObj.silentOff TF.isSilent = tmObj.isSilent TF.setSilent = tmObj.setSilent TF.info = tmObj.info TF.warning = tmObj.warning TF.error = tmObj.error TF.cache = tmObj.cache TF.reset = tmObj.reset TF.indent = tmObj.indent """All messages produced during the feature loading process. It also shows the messages that have been suppressed due to the `silent` parameter. """ TF.ensureLoaded = self.ensureLoaded TF.makeAvailableIn = self.makeAvailableIn TF.footprint = self.footprint setattr(self, "FeatureString", self.Fs) setattr(self, "EdgeString", self.Es) setattr(self, "ComputedString", self.Cs) setattr(self, "AllFeatures", self.Fall) setattr(self, "AllEdges", self.Eall) setattr(self, "AllComputeds", self.Call) setattr(self, "loadLog", self.isLoaded) def Fs(self, fName, warn=True): """Get the node feature sub API. If feature name is not a valid python identifier, or if you do not know its name in advance, you can not use `F.feature`, but you should use `Fs(feature)`. Parameters ---------- fName: string The name of the feature. warn: boolean, optional `True` Whether to warn if the feature is not loaded. Returns ------- The feature API, or `None` if the feature is not loaded. """ if not hasattr(self.F, fName): if warn: self.TF.error(f'Node feature "{fName}" not loaded') return None return getattr(self.F, fName) def Es(self, fName, warn=True): """Get the edge feature sub API. If feature name is not a valid python identifier, or if you do not know its name in advance, you can not use `E.feature`, but you should use `Es(feature)`. Parameters ---------- fName: string The name of the feature. warn: boolean, optional `True` Whether to warn if the feature is not loaded. Returns ------- The feature API, or `None` if the feature is not loaded. """ if not hasattr(self.E, fName): if warn: self.TF.error(f'Edge feature "{fName}" not loaded') return None return getattr(self.E, fName) def Cs(self, fName, warn=True): """Get the computed data sub API. If component name is not a valid python identifier, or if you do not know its name in advance, you can not use `C.component`, but you should use `Cs(component)`. Parameters ---------- fName: string The name of the feature. warn: boolean, optional `True` Whether to warn if the feature is not loaded. Returns ------- The feature API, or `None` if the feature is not loaded. """ if not hasattr(self.C, fName): if warn: self.TF.error(f'Computed feature "{fName}" not loaded') return None return getattr(self.C, fName) def Fall(self, warp=True): """Returns a sorted list of all usable, loaded node feature names. Parameters ---------- warp: boolean, optional True Whether to include the warp features, i.e. `otype`. """ return sorted(x[0] for x in self.F.__dict__.items() if warp or x[0] != OTYPE) def Eall(self, warp=True): """Returns a sorted list of all usable, loaded edge feature names. Parameters ---------- warp: boolean, optional True Whether to include the warp features, i.e. `oslots`. """ return sorted(x[0] for x in self.E.__dict__.items() if warp or x[0] != OSLOTS) def Call(self): """Returns a sorted list of all usable, loaded computed data names.""" return sorted(x[0] for x in self.C.__dict__.items()) def isLoaded( self, features=None, pretty=True, valueType=True, path=False, meta="description" ): """Show information about loaded features. Parameters ---------- features: iterable | string, optional None The features to get info for. If absent or None: all features seen by TF. If a string, it is a comma and/or space spearated list of feature names. Otherwise the items of the iterable are feature names. pretty: boolean, optional True If True, it prints an overview of all features seen by TF with information about kind, type, source location and loaded status. The amount of information printed can be tweaked by other parameters. Otherwise, it returns this information as a dict. valueType: boolean, optional True Only relevant if `pretty=True`: whether to print the value type of the values in the feature file. path: boolean, optional True Only relevant if `pretty=True`: whether to print the path name of the feature file. meta: string|list|boolean, optional "description" Only relevant if `pretty=True`: controls what metadata of the feature should be printed. If it is None, False, or the empty string or empty list: no metadata will be printed. It it is the boolean value True: all metadata will be printed. If it is a list of key names or a string with key names separated by white space and/or commas, only these metadata keys will be printed. Returns ------- dict of dict The features are keys, the value per feature is None or a dict with the following information: `None` if the feature is not loaded. If the feature is loaded: * `kind`: `node`, `edge`, `config`, `computed`; * `type` is the type of values: `int`, or `str` or `""`; * `edgeValues`: if an edge feature it indicates whether the edges have values. Otherwise `None`. * `meta`: dictionary containing the metadata of the feature If `pretty`, nothing is returned, but the dict is pretty printed. """ fNames = list(self.TF.features) if features is None else fitemize(features) info = {} for fName in fNames: fMeta = {} fType = None edgeValues = None fSource = None hasInfo = True if fName in self.TF.features: fObj = self.TF.features[fName] fSource = ux(fObj.dirName) fMeta = fObj.metaData fType = fMeta.get("valueType", "") fMeta = {k: v for (k, v) in fMeta.items() if k != "valueType"} isLoadedF = hasattr(self.F, fName) isLoadedE = hasattr(self.E, fName) if isLoadedF or isLoadedE: if isLoadedF: fKind = "node" elif isLoadedE: fKind = "edge" flObj = getattr(self.E, fName) edgeValues = False if fName == "oslots" else flObj.doValues elif ( fName.startswith("__") and fName.endswith("__") and hasattr(self.C, fName.strip("_")) ): fKind = "computed" elif fName in self.TF.features: if fObj.isConfig: fKind = "config" else: hasInfo = False else: hasInfo = False info[fName] = ( dict( kind=fKind, type=fType, meta=fMeta, source=fSource, edgeValues=edgeValues, ) if hasInfo else None ) if pretty: for (fName, fInfo) in sorted(info.items()): if fInfo is None: kind = "NOT LOADED" kind = f" {kind:<10}" fSource = "" metaRep = "" heading = f"{fName:<20}{kind}{fSource}" else: fKind = fInfo["kind"] fMeta = fInfo.get("meta", {}) fType = fInfo.get("type", "") fSource = fInfo.get("source", "") if path else "" fSource = f" {fSource}" if fSource else "" fEV = fInfo.get("edgeValues", "") if valueType: kind = ( f"node ({fType})" if fKind == "node" else f"edge ({fType})" if fKind == "edge" and fEV else "edge" if fKind == "edge" else f"{fKind}" ) kind = f" {kind:<10}" if kind else "" else: kind = "" if meta is True: metaKeys = sorted(fMeta.keys()) metaInfo = fMeta elif not meta: metaInfo = {} else: metaKeys = fitemize(meta) metaInfo = {k: fMeta[k] for k in metaKeys if k in fMeta} heading = f"{fName:<20}{kind}{fSource}" metaRep = "" indent = " " * (len(heading) + 1) if metaInfo: if len(metaKeys) == 1: value = metaInfo.get(metaKeys[0], "") value = "\n".join( wrap(value, width=80, subsequent_indent=indent) ) metaRep = f" {value}" if value else "" else: indent = " " * 21 for k in metaKeys: value = metaInfo.get(k, "") value = "\n".join( wrap( value, width=80, subsequent_indent=f"\t{indent} ", ) ) metaRep += f"\n\t{k:<20} = {value}" msg = f"{heading}{metaRep}" print(msg) return None return info def makeAvailableIn(self, scope): """Exports members of the API to the global namespace. Only the members whose names start with a capital are exported. If you are working with a single data source in your program, it is a bit tedious to write the initial `TF.api.` or `A.api` all the time. By this method you can avoid that. !!! explanation "Longer names" There are also longer names which can be used as aliases to the single capital letters. This might or might not improve the readability of your program. short name | long name --- | --- `N` | `Nodes` `F` | `Feature` `Fs` | `FeatureString` `Fall` | `AllFeatures` `E` | `Edge` `Es` | `EdgeString` `Eall` `AllEdges` `C` | `Computed` `Cs` `ComputedString` `Call` | `AllComputeds` `L` | `Locality` `T` | `Text` `S` | `Search` Parameters ---------- scope: dict A dictionary into which the members of the core API will be inserted. The only sensible choice is: `globals()`. Returns ------- tuple A grouped list of API members that has been hoisted to the global scope. Notes ----- !!! explanation "Why pass `globals()`?" Although we know it should always be `globals()`, we cannot define a function that looks into the `globals()` of its caller. So we have to pass it on. """ for member in dir(self): if "_" not in member and member[0].isupper(): scope[member] = getattr(self, member) if member not in API_REFS: console(f'WARNING: API member "{member}" not documented') grouped = {} for (member, (head, sub, ref)) in API_REFS.items(): grouped.setdefault(ref, {}).setdefault((head, sub), []).append(member) # grouped # node-features=>(Features, node)=>[F, ...] docs = [] for (ref, groups) in sorted(grouped.items()): chunks = [] for ((head, sub), members) in sorted(groups.items()): chunks.append(" ".join(sorted(members, key=lambda x: (len(x), x)))) docs.append((head, ref, tuple(chunks))) return docs # docs # (Features, node-features, ('F ...', ...)) def ensureLoaded(self, features): """Checks if features are loaded and if not loads them. All features in question will be made available to the core API. Parameters ---------- features: string | iterable of strings It is a string containing space separated feature names, or an iterable of feature names. The feature names are just the names of `.tf` files without directory information and without extension. Returns ------- set The names of the features in question as a set of strings. """ F = self.F E = self.E TF = self.TF warning = TF.warning needToLoad = set() loadedFeatures = set() for fName in sorted(flattenToSet(features)): fObj = TF.features.get(fName, None) if not fObj: warning(f'Cannot load feature "{fName}": not in dataset') continue if fObj.dataLoaded and (hasattr(F, fName) or hasattr(E, fName)): loadedFeatures.add(fName) else: needToLoad.add(fName) if len(needToLoad): TF.load( needToLoad, add=True, silent=DEEP, ) loadedFeatures |= needToLoad return loadedFeatures def footprint(self, recompute=False, bySize=True): """Computes the memory footprint in RAM of the loaded TF data. This includes the precomputed data. Parameters ---------- recompute: boolean, optional False The function looks first for earlier computed size data. If that is found, it will be used, and no size computation will take place. Unless this parameter is True. If no earlier computed size data is found, sizes will be computed anyway. bySize: boolean, optional True Whether to sort the features by the size they occupy in RAM. If False, the features will be sorted alphabetically. """ if hasattr(self, "sizes") and not recompute: sizes = self.sizes else: TF = self.TF features = TF.features nFeatures = len(features) sizes = {} for ft in sorted(features): console(f"\rcomputing size of {ft:<30}", newline=False) data = features[ft].data if data is None: continue nData = len(data) sData = deepSize(data) sizes[ft] = (nData, sData) console(f'\r{"":>40}', newline=False) self.sizes = sizes material = "" nFeatures = len(sizes) totals = collections.Counter() for (ft, (nData, sData)) in sorted( sizes.items(), key=(lambda x: (-x[1][1], x[0])) if bySize else lambda x: x[0], ): material += f"{ft} | {nData:,} | {sData:,}\n" totals["nData"] += nData totals["sData"] += sData material += f'TOTAL | {totals["nData"]:,} | {totals["sData"]:,}' header = dedent( f""" # {nFeatures} features feature | members | size in bytes --- | --- | --- """ ) dm(header + material)Instance variables
var ignored-
Which features were found but ignored.
Features are ignored if the feature is also present in another location that is loaded later.
Methods
def Call(self)-
Returns a sorted list of all usable, loaded computed data names.
Expand source code Browse git
def Call(self): """Returns a sorted list of all usable, loaded computed data names.""" return sorted(x[0] for x in self.C.__dict__.items()) def Cs(self, fName, warn=True)-
Get the computed data sub API.
If component name is not a valid python identifier, or if you do not know its name in advance, you can not use
C.component, but you should useCs(component).Parameters
fName:string- The name of the feature.
warn:boolean, optionalTrue- Whether to warn if the feature is not loaded.
Returns
The feature API, or
Noneif the feature is not loaded.Expand source code Browse git
def Cs(self, fName, warn=True): """Get the computed data sub API. If component name is not a valid python identifier, or if you do not know its name in advance, you can not use `C.component`, but you should use `Cs(component)`. Parameters ---------- fName: string The name of the feature. warn: boolean, optional `True` Whether to warn if the feature is not loaded. Returns ------- The feature API, or `None` if the feature is not loaded. """ if not hasattr(self.C, fName): if warn: self.TF.error(f'Computed feature "{fName}" not loaded') return None return getattr(self.C, fName) def Eall(self, warp=True)-
Returns a sorted list of all usable, loaded edge feature names.
Parameters
warp:boolean, optionalTrue- Whether to include the warp features, i.e.
oslots.
Expand source code Browse git
def Eall(self, warp=True): """Returns a sorted list of all usable, loaded edge feature names. Parameters ---------- warp: boolean, optional True Whether to include the warp features, i.e. `oslots`. """ return sorted(x[0] for x in self.E.__dict__.items() if warp or x[0] != OSLOTS) def Es(self, fName, warn=True)-
Get the edge feature sub API.
If feature name is not a valid python identifier, or if you do not know its name in advance, you can not use
E.feature, but you should useEs(feature).Parameters
fName:string- The name of the feature.
warn:boolean, optionalTrue- Whether to warn if the feature is not loaded.
Returns
The feature API, or
Noneif the feature is not loaded.Expand source code Browse git
def Es(self, fName, warn=True): """Get the edge feature sub API. If feature name is not a valid python identifier, or if you do not know its name in advance, you can not use `E.feature`, but you should use `Es(feature)`. Parameters ---------- fName: string The name of the feature. warn: boolean, optional `True` Whether to warn if the feature is not loaded. Returns ------- The feature API, or `None` if the feature is not loaded. """ if not hasattr(self.E, fName): if warn: self.TF.error(f'Edge feature "{fName}" not loaded') return None return getattr(self.E, fName) def Fall(self, warp=True)-
Returns a sorted list of all usable, loaded node feature names.
Parameters
warp:boolean, optionalTrue- Whether to include the warp features, i.e.
otype.
Expand source code Browse git
def Fall(self, warp=True): """Returns a sorted list of all usable, loaded node feature names. Parameters ---------- warp: boolean, optional True Whether to include the warp features, i.e. `otype`. """ return sorted(x[0] for x in self.F.__dict__.items() if warp or x[0] != OTYPE) def Fs(self, fName, warn=True)-
Get the node feature sub API.
If feature name is not a valid python identifier, or if you do not know its name in advance, you can not use
F.feature, but you should useFs(feature).Parameters
fName:string- The name of the feature.
warn:boolean, optionalTrue- Whether to warn if the feature is not loaded.
Returns
The feature API, or
Noneif the feature is not loaded.Expand source code Browse git
def Fs(self, fName, warn=True): """Get the node feature sub API. If feature name is not a valid python identifier, or if you do not know its name in advance, you can not use `F.feature`, but you should use `Fs(feature)`. Parameters ---------- fName: string The name of the feature. warn: boolean, optional `True` Whether to warn if the feature is not loaded. Returns ------- The feature API, or `None` if the feature is not loaded. """ if not hasattr(self.F, fName): if warn: self.TF.error(f'Node feature "{fName}" not loaded') return None return getattr(self.F, fName) def ensureLoaded(self, features)-
Checks if features are loaded and if not loads them.
All features in question will be made available to the core API.
Parameters
features:string | iterableofstrings- It is a string containing space separated feature names,
or an iterable of feature names.
The feature names are just the names of
.tffiles without directory information and without extension.
Returns
set- The names of the features in question as a set of strings.
Expand source code Browse git
def ensureLoaded(self, features): """Checks if features are loaded and if not loads them. All features in question will be made available to the core API. Parameters ---------- features: string | iterable of strings It is a string containing space separated feature names, or an iterable of feature names. The feature names are just the names of `.tf` files without directory information and without extension. Returns ------- set The names of the features in question as a set of strings. """ F = self.F E = self.E TF = self.TF warning = TF.warning needToLoad = set() loadedFeatures = set() for fName in sorted(flattenToSet(features)): fObj = TF.features.get(fName, None) if not fObj: warning(f'Cannot load feature "{fName}": not in dataset') continue if fObj.dataLoaded and (hasattr(F, fName) or hasattr(E, fName)): loadedFeatures.add(fName) else: needToLoad.add(fName) if len(needToLoad): TF.load( needToLoad, add=True, silent=DEEP, ) loadedFeatures |= needToLoad return loadedFeatures def footprint(self, recompute=False, bySize=True)-
Computes the memory footprint in RAM of the loaded TF data.
This includes the precomputed data.
Parameters
recompute:boolean, optionalFalse- The function looks first for earlier computed size data. If that is found, it will be used, and no size computation will take place. Unless this parameter is True. If no earlier computed size data is found, sizes will be computed anyway.
bySize:boolean, optionalTrue- Whether to sort the features by the size they occupy in RAM. If False, the features will be sorted alphabetically.
Expand source code Browse git
def footprint(self, recompute=False, bySize=True): """Computes the memory footprint in RAM of the loaded TF data. This includes the precomputed data. Parameters ---------- recompute: boolean, optional False The function looks first for earlier computed size data. If that is found, it will be used, and no size computation will take place. Unless this parameter is True. If no earlier computed size data is found, sizes will be computed anyway. bySize: boolean, optional True Whether to sort the features by the size they occupy in RAM. If False, the features will be sorted alphabetically. """ if hasattr(self, "sizes") and not recompute: sizes = self.sizes else: TF = self.TF features = TF.features nFeatures = len(features) sizes = {} for ft in sorted(features): console(f"\rcomputing size of {ft:<30}", newline=False) data = features[ft].data if data is None: continue nData = len(data) sData = deepSize(data) sizes[ft] = (nData, sData) console(f'\r{"":>40}', newline=False) self.sizes = sizes material = "" nFeatures = len(sizes) totals = collections.Counter() for (ft, (nData, sData)) in sorted( sizes.items(), key=(lambda x: (-x[1][1], x[0])) if bySize else lambda x: x[0], ): material += f"{ft} | {nData:,} | {sData:,}\n" totals["nData"] += nData totals["sData"] += sData material += f'TOTAL | {totals["nData"]:,} | {totals["sData"]:,}' header = dedent( f""" # {nFeatures} features feature | members | size in bytes --- | --- | --- """ ) dm(header + material) def isLoaded(self, features=None, pretty=True, valueType=True, path=False, meta='description')-
Show information about loaded features.
Parameters
features:iterable | string, optionalNone- The features to get info for. If absent or None: all features seen by TF. If a string, it is a comma and/or space spearated list of feature names. Otherwise the items of the iterable are feature names.
pretty:boolean, optionalTrue- If True, it prints an overview of all features seen by TF with information about kind, type, source location and loaded status. The amount of information printed can be tweaked by other parameters. Otherwise, it returns this information as a dict.
valueType:boolean, optionalTrue- Only relevant if
pretty=True: whether to print the value type of the values in the feature file. path:boolean, optionalTrue- Only relevant if
pretty=True: whether to print the path name of the feature file. meta:string|list|boolean, optional"description"-
Only relevant if
pretty=True: controls what metadata of the feature should be printed.If it is None, False, or the empty string or empty list: no metadata will be printed.
It it is the boolean value True: all metadata will be printed.
If it is a list of key names or a string with key names separated by white space and/or commas, only these metadata keys will be printed.
Returns
dictofdict-
The features are keys, the value per feature is None or a dict with the following information:
Noneif the feature is not loaded.If the feature is loaded:
kind:node,edge,config,computed;typeis the type of values:int, orstror"";edgeValues: if an edge feature it indicates whether the edges have values. OtherwiseNone.meta: dictionary containing the metadata of the feature
If
pretty, nothing is returned, but the dict is pretty printed.
Expand source code Browse git
def isLoaded( self, features=None, pretty=True, valueType=True, path=False, meta="description" ): """Show information about loaded features. Parameters ---------- features: iterable | string, optional None The features to get info for. If absent or None: all features seen by TF. If a string, it is a comma and/or space spearated list of feature names. Otherwise the items of the iterable are feature names. pretty: boolean, optional True If True, it prints an overview of all features seen by TF with information about kind, type, source location and loaded status. The amount of information printed can be tweaked by other parameters. Otherwise, it returns this information as a dict. valueType: boolean, optional True Only relevant if `pretty=True`: whether to print the value type of the values in the feature file. path: boolean, optional True Only relevant if `pretty=True`: whether to print the path name of the feature file. meta: string|list|boolean, optional "description" Only relevant if `pretty=True`: controls what metadata of the feature should be printed. If it is None, False, or the empty string or empty list: no metadata will be printed. It it is the boolean value True: all metadata will be printed. If it is a list of key names or a string with key names separated by white space and/or commas, only these metadata keys will be printed. Returns ------- dict of dict The features are keys, the value per feature is None or a dict with the following information: `None` if the feature is not loaded. If the feature is loaded: * `kind`: `node`, `edge`, `config`, `computed`; * `type` is the type of values: `int`, or `str` or `""`; * `edgeValues`: if an edge feature it indicates whether the edges have values. Otherwise `None`. * `meta`: dictionary containing the metadata of the feature If `pretty`, nothing is returned, but the dict is pretty printed. """ fNames = list(self.TF.features) if features is None else fitemize(features) info = {} for fName in fNames: fMeta = {} fType = None edgeValues = None fSource = None hasInfo = True if fName in self.TF.features: fObj = self.TF.features[fName] fSource = ux(fObj.dirName) fMeta = fObj.metaData fType = fMeta.get("valueType", "") fMeta = {k: v for (k, v) in fMeta.items() if k != "valueType"} isLoadedF = hasattr(self.F, fName) isLoadedE = hasattr(self.E, fName) if isLoadedF or isLoadedE: if isLoadedF: fKind = "node" elif isLoadedE: fKind = "edge" flObj = getattr(self.E, fName) edgeValues = False if fName == "oslots" else flObj.doValues elif ( fName.startswith("__") and fName.endswith("__") and hasattr(self.C, fName.strip("_")) ): fKind = "computed" elif fName in self.TF.features: if fObj.isConfig: fKind = "config" else: hasInfo = False else: hasInfo = False info[fName] = ( dict( kind=fKind, type=fType, meta=fMeta, source=fSource, edgeValues=edgeValues, ) if hasInfo else None ) if pretty: for (fName, fInfo) in sorted(info.items()): if fInfo is None: kind = "NOT LOADED" kind = f" {kind:<10}" fSource = "" metaRep = "" heading = f"{fName:<20}{kind}{fSource}" else: fKind = fInfo["kind"] fMeta = fInfo.get("meta", {}) fType = fInfo.get("type", "") fSource = fInfo.get("source", "") if path else "" fSource = f" {fSource}" if fSource else "" fEV = fInfo.get("edgeValues", "") if valueType: kind = ( f"node ({fType})" if fKind == "node" else f"edge ({fType})" if fKind == "edge" and fEV else "edge" if fKind == "edge" else f"{fKind}" ) kind = f" {kind:<10}" if kind else "" else: kind = "" if meta is True: metaKeys = sorted(fMeta.keys()) metaInfo = fMeta elif not meta: metaInfo = {} else: metaKeys = fitemize(meta) metaInfo = {k: fMeta[k] for k in metaKeys if k in fMeta} heading = f"{fName:<20}{kind}{fSource}" metaRep = "" indent = " " * (len(heading) + 1) if metaInfo: if len(metaKeys) == 1: value = metaInfo.get(metaKeys[0], "") value = "\n".join( wrap(value, width=80, subsequent_indent=indent) ) metaRep = f" {value}" if value else "" else: indent = " " * 21 for k in metaKeys: value = metaInfo.get(k, "") value = "\n".join( wrap( value, width=80, subsequent_indent=f"\t{indent} ", ) ) metaRep += f"\n\t{k:<20} = {value}" msg = f"{heading}{metaRep}" print(msg) return None return info def makeAvailableIn(self, scope)-
Exports members of the API to the global namespace.
Only the members whose names start with a capital are exported.
If you are working with a single data source in your program, it is a bit tedious to write the initial
TF.api.orA.apiall the time. By this method you can avoid that.Longer names
There are also longer names which can be used as aliases to the single capital letters. This might or might not improve the readability of your program.
short name long name NNodesFFeatureFsFeatureStringFallAllFeaturesEEdgeEsEdgeStringEallAllEdgesCComputedCsComputedStringCallAllComputedsLLocalityTTextSSearchParameters
scope:dict- A dictionary into which the members of the core API will be inserted.
The only sensible choice is:
globals().
Returns
tuple- A grouped list of API members that has been hoisted to the global scope.
Notes
Why pass
globals()?Although we know it should always be
globals(), we cannot define a function that looks into theglobals()of its caller. So we have to pass it on.Expand source code Browse git
def makeAvailableIn(self, scope): """Exports members of the API to the global namespace. Only the members whose names start with a capital are exported. If you are working with a single data source in your program, it is a bit tedious to write the initial `TF.api.` or `A.api` all the time. By this method you can avoid that. !!! explanation "Longer names" There are also longer names which can be used as aliases to the single capital letters. This might or might not improve the readability of your program. short name | long name --- | --- `N` | `Nodes` `F` | `Feature` `Fs` | `FeatureString` `Fall` | `AllFeatures` `E` | `Edge` `Es` | `EdgeString` `Eall` `AllEdges` `C` | `Computed` `Cs` `ComputedString` `Call` | `AllComputeds` `L` | `Locality` `T` | `Text` `S` | `Search` Parameters ---------- scope: dict A dictionary into which the members of the core API will be inserted. The only sensible choice is: `globals()`. Returns ------- tuple A grouped list of API members that has been hoisted to the global scope. Notes ----- !!! explanation "Why pass `globals()`?" Although we know it should always be `globals()`, we cannot define a function that looks into the `globals()` of its caller. So we have to pass it on. """ for member in dir(self): if "_" not in member and member[0].isupper(): scope[member] = getattr(self, member) if member not in API_REFS: console(f'WARNING: API member "{member}" not documented') grouped = {} for (member, (head, sub, ref)) in API_REFS.items(): grouped.setdefault(ref, {}).setdefault((head, sub), []).append(member) # grouped # node-features=>(Features, node)=>[F, ...] docs = [] for (ref, groups) in sorted(grouped.items()): chunks = [] for ((head, sub), members) in sorted(groups.items()): chunks.append(" ".join(sorted(members, key=lambda x: (len(x), x)))) docs.append((head, ref, tuple(chunks))) return docs