I am trying to store some extra information with DataFrames directly in the same DataFrame, such as some parameters describing the data stored.
I added this information just as extra attributes to the DataFrame:
df.data_origin = 'my_origin'
print(df.data_origin)
But when it is saved and loaded, those extra attributes are lost:
df.to_pickle('pickle_test.pkl')
df2 = pd.read_pickle('pickle_test.pkl')
print(len(df2))
print(df2.definition)
...
465387
>>> AttributeError: 'DataFrame' object has no attribute 'definition'
The workaround I have found is to save the dict of the DataFrame and then assign it to the dict of an empty DataFrame:
with open('modified_dataframe.pkl', "wb") as pkl_out:
pickle.dump(df.__dict__, pkl_out)
df2 = pd.DataFrame()
with open('modified_dataframe.pkl', "rb") as pkl_in:
df2.__dict__ = pickle.load(pkl_in)
print(len(df2))
print(df2.data_origin)
...
465387
my_origin
It seems to work, but:
Is there a better way to do it?
Am I losing information? (apparently, all the data is there)
Here a different solution is discussed, but I would like to know if the approach of saving the dict of a class is valid to hold its entire information.
EDIT: Ok, I found the big drawback. This works fine to save single DataFrames in isolated files, but will not work if I have dictionaries, lists or similar with DataFrames in them.
I suggest that you can get your things done by making a new child class for pandas.DataFrame, make a new class inherit things from pandas.DataFrame class, and add your wanted attributes there. This may seem a bit spooky, but you can play around with it safely when you using in different places. Other stuff might be useful for specific cases though.
Related
I've been struggling with this matter for 2 full days now due to my incompetence. After trying almost all stackoverflow and other solutions I could find sadly still no luck.
I'm using Tabular-Py to import tables from PDFs. After which it's already "perfectly" in what seems to be a dataframe. The part of the code used for this is:
tables = tabula.read_pdf(file, pages=18, lattice=True, multiple_tables = False)
Print(Tables)
[Output after printing the table]
[1]: https://i.stack.imgur.com/82Qpa.png
However, it seems to be a list object, as it's blocking me from doing anything else with it besides printing. Even using integers and renaming columns doesn't work due to the errors leading back to "Cannot XX because it's a list object". I was under the impression Tabular makes a direct Pandas Dataframe.
Now when I try to add the following code to rename the columns as desired:
tables.columns = ['HS_Code', 'Product', 'PreviousMonth', 'CurrentMonth', 'LastYear']
I get the error :
AttributeError: 'list' object has no attribute 'columns'
I've tried many forms of renaming and using different sets of output such as Json. Still no luck, it's still a "list object".
Does anyone have experience with this matter? How can I ensure the Table/Dataframe I have is an actual dataframe instead of a list object?
Any tips would be highly appreciated.
I am not familiar with tabula-py objects but considering this post you can do the following:
use pandas.read_clipboard() after copying the pdf content by hand
or 2. save the tabula-py object as csv and use pandas.read_csv() to get the DataFrame
Afterwards you are able to manipulate the data (e.g. change column names) using pandas.
I want to get the discord.user_id, I am VERY new to python and just need help getting this data.
I have tried everything and there is no clear answer online.
currently, this works to get a data point in the attributes section
pledge.relationship('patron').attribute('first_name')
You should try this :
import pandas as pd
df = pd.read_json(path_to_your/file.json)
The ourput will be a DataFrame which is a matrix, in which the json attributes will be the names of the columns. You will have to manipulate it afterwards, which is preferable, as the operations on DataFrames are optimized in terms of processing time.
Here is the official documentation, take a look.
Assuming the whole object is call myObject, you can obtain the discord.user_id by calling myObject.json_data.attributes.social_connections.discord.user_id
my code is on the bottom
"parse_xml" function can transfer a xml file to a df, for example, "df=parse_XML("example.xml", lst_level2_tags)" works
but as I want to save to several dfs so I want to have names like df_ first_level_tag, etc
when I run the bottom code, I get an error "f'df_{first_level_tag}'=parse_XML("example.xml", lst_level2_tags)
^
SyntaxError: can't assign to literal"
I also tried .format method instead of f-string but it also hasn't worked
there are at least 30 dfs to save and I don't want to do it one by one. always succeeded with f-string in Python outside pandas though
Is the problem here about f-string/format method or my code has other logic problem?
if necessary for you, the parse_xml function is directly from this link
the function definition
for first_level_tag in first_level_tags:
lst_level2_tags = []
for subchild in root[0]:
lst_level2_tags.append(subchild.tag)
f'df_{first_level_tag}'=parse_XML("example.xml", lst_level2_tags)
This seems like a situation where you'd be best served by putting them into a dictionary:
dfs = {}
for first_level_tag in first_level_tags:
lst_level2_tags = []
for subchild in root[0]:
lst_level2_tags.append(subchild.tag)
dfs[first_level_tag] = parse_XML("example.xml", lst_level2_tags)
There's nothing structurally wrong with your f-string, but you generally can't get dynamic variable names in Python without doing ugly things. In general, storing the values in a dictionary ends up being a much cleaner solution when you want something like that.
One advantage of working with them this way is that you can then just iterate over the dictionary later on if you want to do something to each of them. For example, if you wanted to write each of them to disk as a CSV with a name matching the tag, you could do something like:
for key, df in dfs.items():
df.to_csv(f'{key}.csv')
You can also just refer to them individually (so if there was a tag named a, you could refer to dfs['a'] to access it in your code later).
I have a dictionary that is being built while iterating through objects. Now same object can be accessed multiple times. And I'm using object itself as a key.
So if same object is accessed more than once, then key becomes not unique and my dictionary is no longer correct.
Though I need to access it by object, because later on if someone wants access contents by it, they can request to get it by current object. And it will be correct, because it will access the last active object at that time.
So I'm wondering if it is possible to wrap object somehow, so it would keep its state and all attributes the same, but the only difference would be this new kind of object which is actually unique.
For example:
dct = {}
for obj in some_objects_lst:
# Well this kind of wraps it, but it loses state, so if I would
# instantiate I would lose all information that was in that obj.
wrapped = type('Wrapped', (type(obj),), {})
dct[wrapped] = # add some content
Now if there are some better alternatives than this, I would like to hear it too.
P.S. objects being iterated would be in different context, so even if object is the same, it would be treated differently.
Update
As requested, to give better example where the problem comes from:
I have this excel reports generator module. Using it, you can generate various excel reports. For that you need to write configuration using python dictionary.
Now before report is generated, it must do two things. Get metadata (metadata here is position of each cell that will be when report is about to be created) and second, parse configuration to fill cells with content.
One of the value types that can be used in this module, is formula (excel formulas). And the problem in my question is specifically with one of the ways formula can be computed: formula values that are retrieved for parent , that are in their childs.
For example imagine this excel file structure:
A | B | C
Total Childs Name Amount
1 sum(childs)
2 child_1 10
3 child_2 20
4 sum(childs)
...
Now in this example sum on cell 1A, would need to be 10+20=30 if sum would use expression to sum their childs column (in this case C column). And all of this is working until same object (I call it iterables) is repeated. Because when building metadata it I need to store it, to retrieve later. And key is object being iterated itself. So now when it will be iterated again when parsing values, it will not see all information, because some will overwritten by same object.
For example imagine there are invoice objects, then there are partner objects which are related with invoices and there are some other arbitrary objects that given invoice and partner produce specific amounts.
So when extracting such information in excel, it goes like this:
inoice1 -> partner1 -> amount_obj1, amount_obj2
invoice2 -> partner1 -> amount_obj3, amount_obj4.
Notice that partner in example is the same. Here is the problem, because I can't store this as key, because when parsing values, I will iterate over this object twice when metadata will actually hold values for amount_obj3 and amount_obj4
P.S Don't know if I explained it better, cause there is lots of code and I don't want to put huge walls of code here.
Update2
I'll try to explain this problem from more abstract angle, because it seems being too specific just confuses everyone even more.
So given objects list and empty dictionary, dictionary is built by iterating over objects. Objects act as a key in dictionary. It contains metadata used later on.
Now same list can be iterated again for different purpose. When its done, it needs to access that dictionary values using iterated object (same objects that are keys in that dictionary). But the problem is, if same object was used more than once, it will have only latest stored value for that key.
It means object is not unique key here. But the problem is the only thing I know is the object (when I need to retrieve the value). But because it is same iteration, specific index of iteration will be the same when accessing same object both times.
So uniqueness I guess then is (index, object).
I'm not sure if I understand your problem, so here's two options. If it's object content that matters, keep object copies as a key. Something crude like
new_obj = copy.deepcopy(obj)
dct[new_obj] = whatever_you_need_to_store(new_obj)
If the object doesn't change between the first time it's checked by your code and the next, the operation is just performed the second time with no effect. Not optimal, but probably not a big problem. If it does change, though, you get separate records for old and new ones. For memory saving you will probably want to replace copies with hashes, __str__() method that writes object data or whatever. But that depends on what your object is; maybe hashing will take too much time for miniscule savings in memory. Run some tests and see what works.
If, on the other hand, it's important to keep the same value for the same object, whether the data within it have changed or not (say, object is a user session that can change its data between login and logoff), use object ids. Not the builtin id() function, because if the object gets GCed or deleted, some other object may get its id. Define an id attribute for your objects and make sure different objects cannot possibly get the same one.
I'm a beginner, and the answers I've found online so far for this have been too complicated to be useful, so I'm looking for an answer in vocabulary and complexity similar to this writing.
I'm using python 2.7 in ipython notebook environment, along with related modules as distributed by anaconda, and I need to learn about the library-specific objects in the course of my daily work. The case I'm using here is a pandas dataframe object but the answer must work for any object of python or of an imported module.
I want to be able to print a list of methods for the given object. Directly from my program, in a concise and readable format. Even if it's just the method names in a list by alphabetical order, that would be great. A bit more detail would be even better, an ordering based on what it does is fine, but I'd like the output to look like a table, one row per method, and not big blocks of text. What i've tried is below, and it fails for me because it's unreadable. It puts copies of my data between each line, and it has no formatting.
(I love stackoverflow. I aspire to have enough points someday to upvote all your wonderful answers.)
import pandas
import inspect
data_json = """{"0":{"comment":"I won\'t go to school"}, "1":{"note":"Then you must stay in bed"}}"""
data_df = pandas.io.json.read_json(data_json, typ='frame',
dtype=True, convert_axes=True,
convert_dates=True, keep_default_dates=True,
numpy=False, precise_float=False,
date_unit=None)
inspect.getmembers(data_df, inspect.ismethod)
Thanks,
- Sharon
Create an object of type str:
name = "Fido"
List all its attributes (there are no “methods” in Python) in alphabetical order:
for attr in sorted(dir(name)):
print attr
Get more information about the lower (function) attribute:
print(name.lower.__doc__)
In an interactive session, you can also use the more convenient
help(name.lower)
function.