Extract data from files compressed in gz archives

I need a script that can scan a large number of .gz files as fast as possible and extract certain data from them. I need the script to be able to do this without having to decompress all the files (there are over 40000 of them!) Speed is critical and I would prefer a solution in python but am not closed to the idea of using other languages especially if it will run faster. Also, the source files should not be modified just checking them for matches and writing the results to a "results" file. Every .gz file contains a single json file containing information on up to 750 different keywords/items so it needs to be able to recurse through the file and match the data found with the correct keyword. Here is an example: The data I need to locate is entries in the json file that include the "images" tag. Here is a sample: { "date": "2016-06-10", "gl": "us", "hl": "en", "custom_id": "2", "keyword": "cat drinking water", "data": { "1": { "pos": 1, "href": "[login to view URL]", "title": "", "description": "", "tags": [ "images" ] }, "2": { "pos": 2, "href": "[login to view URL]", "title": "Amazing Slow Motion Cat Drinking - YouTube", "description": "Discover the beauty of a cat in super slow motion thanks to a high definition ... Giant 6ft Water Balloon - The ...", "tags": [ "video" ] }, "3": { "pos": 3, "href": "[login to view URL]", "title": "Problems With a Cat Drinking Excessive Water - Pets", "description": "Perhaps better known as finicky eaters, cats aren't prolific water drinkers. If your cat is drinking a lot of water , it could be a sign of a serious health issue including \u00a0...", "tags": [] ... As you can see, the first entry under "data" has the "images" tag. Once the "images" tag is found I need to get this information from the "top" of the entry as seen in the example above: "gl": "us", "hl": "en", "custom_id": "2", "keyword": "cat drinking water", This information should be saved to a text file in this format: "GL;HL;ID;keyword" So based on this example it would save this to the file: "US;EN;2;cat drinking water" For every match found in all the .gz files a new line should be written to the "results" file so everything is stored in a single file. I have attached a sample .gz file you can test with. I need this completes as soon as possible so how soon you can begin and complete work will factor into my bid selection. Thanks and feel free to ask questions!