吳若準旣聞顧廣圻改定洛陽伽藍記本注之說於其舅，而有「集證」之作。唯吳本正文太簡，子注過繁，其所分析疑與楊記舊觀相去甚遠，唐晏因是有「洛陽伽藍記鉤沈」之作。晏本滿州人，姓瓜爾佳氏，淸社旣屋，始改氏唐，光緒八年（一八八二）進士，精於詞章書畫金石考據之學。民國四年（一九一五）避居滬上，因嫌吳氏集證『限域未淸，混淆不免，雖少勝於舊編，猶未盡夫塵障』，乃『信手鈎乙，數則以後，迎刄而解，都已盡卷。』成此「鉤沈」一編。唐氏此書全用水經注體裁，涉及寺事市里第宅，例高一格書；附注概低一格；衒之案語，則作夾行。誤字意定，以△別之；考訂之字，旁以圈別之。又衒之此書，意在借伽藍而存時事，故搜採「魏書」、「北史」諸書，於記中人物，皆略注大概，而它有足以引證此書者，亦附焉。吳氏集證附圖，亦為改訂，仍冠篇首。唯就唐氏凡例案之，亦不無可議者，張宗祥「合校本後記」云：（鉤沈本）第五卷原本注文，且誤入正文，則亦未為盡合也。』「續修四庫全書提要」更擧數例以證其誤：如卷一永寧寺浮圖一節之「衒之常與河南尹胡孝世共登之下臨雲雨信哉不虛」等二十一字，應作夾行，而唐本誤為注文。又如卷五禪虛寺條「所謂永平里也注卽漢太上王廣處」，唐本列為正文，張氏則以為『此處「注」字幸存，「卽漢太上王廣處」六字（彬案：應為七字），明係注文，不得誤入正文。』諸如此例，可議尚多。要之衒之此書，本文子注不易別析，苟無如隱以前之古本，强為分析，終不免於千慮之失耳。唐氏鉤沈上繼吳氏集證而作，下啟張氏合校本出，因據民國六年潮陽鄭氏刋本景印。因係僅供學者比閱，故「集證」與「鉤沈」兩本，皆以四合一縮印之。
After Wu Ruozhun heard from his uncle about Gu Guangqi's revised edition of *A Record of Buddhist Monasteries in Luoyang* (Luoyang Qielan Ji), he produced the *Collected Annotations*. However, Wu's main text was too brief, and his sub-commentary excessively lengthy; his analysis likely strayed far from Yang Xuanzhi's original structure. Thus, Tang Yan wrote *A Recovery of the Luoyang Qielan Ji*. Tang was a Manchu of the Guwalgiya clan. After the Qing dynasty fell, he changed his surname to Tang. A *jinshi* (metropolitan graduate) in 1882, he excelled in poetry, calligraphy, painting, and epigraphic studies. In 1915, he took refuge in Shanghai. Dissatisfied that Wu’s *Collected Annotations* had "unclear boundaries and unavoidable confusion—though slightly better than older editions, it had not fully cleared the dust of textual error," he "casually began outlining corrections; after a few entries, the work fell into place, and the entire text was completed." Thus, *A Recovery* emerged.
Tang's book fully adopts the format of the *Commentary on the Water Classic* (*Shuijingzhu*). Entries on monasteries, markets, residences, and mansions are printed in a raised line; annotations are in an indented line; Yang Xuanzhi’s own remarks appear as interlinear notes. Suspected errors are marked with ▲; characters identified through textual criticism are circled. Since Yang’s work aimed to preserve contemporary events through monastic records, Tang extensively cites the *Weishu* (Book of Wei) and *Beishi* (History of the Northern Dynasties), briefly annotating figures mentioned in the text and attaching any other material that corroborates it. He also revised Wu’s attached map and placed it at the beginning of the book.
However, even following Tang’s own editorial principles, there are points to criticize. Zhang Zongxiang, in his *Postscript to the Collated Edition*, states: "In the fifth volume of the *Recovery* version, some of the original annotation mistakenly enters the main text—so it is not entirely consistent." The *Continuation of the Complete Library of the Four Treasuries* (*Xuxiu Siku Quanshu*) abstract further cites examples of errors. For instance, the passage on the Yongning Monastery pagoda in volume one: "Xuanzhi often ascended it with Henan Governor Hu Xiaoshi; it rose above the clouds and rain, truly not in vain" (21 characters) should be an interlinear note, but Tang’s edition lists it as annotation. Another example: in the Chanxu Monastery entry in volume five, the phrase "what is called Yongping li, annotation: the site where the Han Emperor of the Upper Realm's father expanded" is listed as main text by Tang. Zhang argues: "The word 'annotation' survives here, and the six characters (Note by Bin: actually seven) ‘the site where the Han Emperor of the Upper Realm’s father expanded’ are clearly a sub-commentary and should not be placed in the main text." Many similar debatable points remain. Ultimately, distinguishing the main text from sub-commentary in Yang’s work is difficult without a pre-Xuanzang ancient version; forced analysis inevitably leads to occasional errors.
Tang’s *Recovery* follows Wu’s *Collected Annotations* and precedes Zhang’s *Collated Edition*. It is reproduced here from the 1917 Chaoyang Zheng family edition, printed as a tetramerous reduction of both the *Collected Annotations* and the *Recovery* versions for scholars’ comparative reading.