Is there a way to see if some important languages are excluded from the mining?

I have seen the language stats report in the link 'Mined Projects'. Are these the 13 most widespread languages and everything else is 'below Kotlin' or there are holes with widespread languages in between?

What's heppening? Hi, when i use the web interface without any filter to save the result it gives me "504 Gateway Time-Out error".

How to see the problem again? Open web interface without choosing anything just clike on Search button and then clicke on Download CSV.

Output Result: Generated url:

https://seart-ghs.si.usi.ch/api/r/download/csv?hasWiki=false&onlyForks=false&hasLicense=false&nameEquals=false&hasPulls=false&excludeForks=false&hasIssues=false

Btw. Can you share the whole up-to-date dataset? I found a dataset on zenodo, but it is not up-to-date

Values for 'Total Issues' and 'Total Pull Reqs' are sometimes 'null' instead of 0.

For example:

Link to GitHub repo: https://github.com/0xc0000054/pdn-gmic

Here are some correct examples:

I recently pulled some data and noticed that there were a few projects that had a reported number of contributors that is quite odd.

I noticed this for three projects in the data set I am using. The projects in question are: radeonopencompute/rock-kernel-driver youling257/android-mainline xanmod/linux

In the csv output generated the number of contributors for these projects is 9.22337203685477E+018. Or at least it is after I import it to a spreadsheet.

On the website search results the number of contributors for these projects is 9223372036854776000.

Based on the number it looks like this is a JavaScript error causing the maximum integer value to be displayed. Obviously the number of contributors to the projects in question cannot be close to these reported values.

Hi, I found the 'id' field misleading, interpreting it as an id to retrieve the project from the GH API. Can it be removed if it's only a GHSearch internal identifier?

Hi, I was wondering if it could be useful to have forked parameters like 'forked from' and 'fork source' (specifically for repos that are forks).

I know that every possible detail might be useful in one or more cases. Just wondering if these two fields are worth having in the output.

Example: if you need to analyze forks and fork chains starting from a subset of projects GHS could be used directly for that kind of research if fork chains could be reconstructed directly without scraping GH.

Hi, I would find very useful to have the search parameters stored in the exported JSON. Not sure about other export formats but I suppose a similar approach would still apply.

Top level in JSON there is 'items'. Adding something like 'search parameters' at the same level could help in keeping the specifics close to the data (instead of relying on file renaming or just manual annotation). It would help in keeping track of the search criteria months after when a dataset pops out from a folder and I can't remember exactly the specifics of it.

Hi, maybe it would be useful to have a statistics breakdown after returning a successful search (i.e., how many different languages, average contributors, min-max-average commits just to name a few examples).

An overview of the selected bunch of repositories could help in refining search criteria or have a first glimpse of what's inside a potential dataset even before exporting the output.

In general min, max, average, median, unique values (and/or other usual descriptive statistics) for each 'applicable' output field could be useful for recap/early analysis.

Hi, I was wondering if it's possible to keep the search parameters after returning from a search ('Back' button).

Example use case:

• Set some filtering criteria
• Search
• Get results
• Want to refine/fine-tune/adjust search criteria
• hit the back button
• modify criteria I want to adjust (others are "saved/restored" automatically)

Maybe can be also done by supporting the browser's back button (actually ignoring 'in site' navigation) and removing the custom one.

Consider searching apache/commons-math (in the exact match mode), and it takes the web app up to 1 minute (!!) to perform the search. This is the same for downloading the result (i.e., a csv with one row).

Hi, when I search for projects in C++ the result on webpage is correct but the result in the download file is not correct. The file seems to only shows projects written in C.

Hi, I had problems in finding a unique identifier resilient to repo name changes, ownership changes, etc.

Can you think of any such identifier that can be exposed in the export to have a more reliable (time-invariant) way of retrieving a repository?

I see from GH rest API that there is indeed an id returned but I'm not sure if this is usable and/or if this is already what you export as id (possibly relate to #16)